Raza Tahir-Kheli

Raza Tahir-Kheli
Ordinary
Differential
Equations
Mathematical Tools for Physicists
Ordinary Differential Equations
Raza Tahir-Kheli
Ordinary Differential
Equations
Mathematical Tools for Physicists
123
Raza Tahir-Kheli
Department of Physics
Temple University
Philadelphia, PA, USA
ISBN 978-3-319-76405-4 ISBN 978-3-319-76406-1 (eBook)

https://doi.org/10.1007/978-3-319-76406-1
Library of Congress Control Number: 2018933457
© Springer Nature Switzerland AG 2018

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made. The publisher remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Dedicated to my friends
Sir Roger J. Elliott, Kt., F.R.S.
(in Memorium)
Sir Anthony J.Leggett, KBE, F.R.S., Nobel
Laureate
Alan J. Heeger, Nobel Laureate
J. Robert Schrieffer, Nobel Laureate
my wife
Ambassador Shirin Raziuddin Tahir-Kheli
and our grandchildren
Taisiya, Alexander, Cyrus, Gladia
Preface
This book is intended as a reader-friendly source for self-study and as an accessible

textbook that contains many solved problems.
My experience teaching at Temple University the mathematical physics course
that includes ordinary differential equations has guided the writing of this textbook.
The mathematical physics course is offered to undergraduates in their pre- or final
year of study in physics, engineering, chemistry, earth and environmental sciences,
or mathematical biology. It is also taken by beginning graduate students working
toward a master’s degree.
Years of teaching have helped me understand what works for students and what
does not. In particular, I have learned that the more attention a student pays to
taking notes, the less he/she understands of the subject matter of the lecture being
delivered. Further, I have noticed that when, a week in advance of the delivery
of the lecture, a student is provided details of the algebra to be used, solutions to the
problems to be discussed, and some brief information about the ideas that are
central to the lecture to be given the following week, it obviates much of the need
for note taking during the delivery of the lecture. Another important experience that
has guided the writing of this textbook is the pedagogical benefit that accrues from
an occasional, quick, recapitulation of the relevant results that have already been
presented in an earlier lecture. All this results in better comprehension of the subject
matter.
Both for the purposes of elucidation of the concepts introduced in the text and
for providing practical problem solving support, solutions to a large number of
examples have been included. Many of the solutions provided contain much greater
detail than would be necessary for presentation in a lecture itself or needed by
teachers or more advanced practitioners of the subject. These solutions are there in
the given form to offer encouragement and support to the student: both for
self-study and to allow for fuller understanding of the subject matter. Therefore, it is
as important for the reader to assimilate the subject matter of the book as it is to
independently work through the solved problems before reading through the
solutions provided. Another notable feature of the book is that equations are
vii
viii Preface
numbered in seriatim. Included in the numbering process is the chapter number. As

a result, access to an equation being referred to is both easy and fast.
In Chap. 1, the differential operator D, is introduced and the rules it follows are
articulated.
A cursory look at the vocabulary is taken in Chap. 2 titled ‘Some Definitions.’
Noted there is the terminology and some of the definitions that are used in this
manuscript. A function involving a dependent variable, say u ; only a single inde-
pendent variable, say x ; and at least one derivative of u with respect to x is an
ordinary differential equation (ODE). Explicit (ODE) and implicit (ODE) are
defined. Generally, explicit (ODEs) are easier to treat than implicit (ODEs).
Therefore, mostly the explicit (ODEs) are treated in this book. A linear ordinary
differential equation is of first degree in the variable, uðxÞ, as well as its derivatives.
Homogeneous linear ordinary differential equation and inhomogeneous linear
ordinary differential equation are defined. When an ODE cannot be expressed in
linear form, it is said to be nonlinear (ODE). Just as implicit (ODE) is harder to
solve than explicit (ODE), solving nonlinear (ODE) requires more effort than
solving linear (ODE). Furthermore, simple treatment of nonlinear (ODE) cannot be
guaranteed to succeed.
Linear ordinary differential equations with known constant coefficients are
treated in Chap. 3. Procedure for solving homogeneous linear ordinary differential
equations with known constant coefficients, the method of undetermined coeffi-
cients, and calculation of the particular integral, Ipi , for inhomogeneous linear
ordinary differential equations are all described in detail. The concepts of linear
independence, linear dependence, and the use of Wronskians are elucidated. And
simultaneous linear ordinary differential equations with constant coefficients are
studied in detail.
In Chap. 4, the analysis is extended to linear ordinary differential equations that
have variable coefficients. Depending upon the nature of the variable coefficient, the
linear ordinary differential equation with variable coefficients can be much harder to
solve than linear ordinary differential equations with constant coefficients. For
simplicity, therefore, only the first-order and first-degree equations of type
Bernoulli equation are treated. Included also is a discussion of equations that can be
transformed into Bernoulli equation.
Chapter 5 deals with Green’s function and Laplace transforms.
In Chap. 6, equations Beyond Bernoulli—to be called ‘Special Type’ differential
equations—are studied. Included there are the Clairaut equations [compare (6.2)–
(6.13)], Lagrange equation [compare (6.19)–(6.31)], the separable equations

[compare (6.32)–(6.35)], and the dy dx ¼ U x equations [compare (6.36)–(6.73)].
y
In addition, there are the so-called exact [compare (6.74)–(6.91)] and inexact
equations [compare (6.92)–(6.241)], Riccati equations [compare (6.242)–(6.268)],
Euler equations [compare (6.269)–(6.315)], and the factorable equations [compare
(6.316)–(6.344)]. Singular solution of Clairaut equation is discussed and calculated
both by an informal procedure and a formal procedure.
Preface ix
Studied in Chap. 7 are equations where ‘Special Situations’ obtain. For instance,
a given differential equation may be integrable. Similarly, there are equations that
have both the independent and the dependent variables missing in any explicit form.
And then, there are those that explicitly contain only the independent variable, or
only the dependent variable. All these equations provide possible routes to suc-
cessful treatment. An interesting special situation, called ‘Order Reduction,’ comes
into play if one of the n non-trivial solutions of an nth order homogeneous linear
ordinary differential equation is already known. Then, the given equation can be
reduced to one of the ðn 1Þth order. Homogeneous linear ordinary second-order
differential equations have been treated before. Studied in this chapter are inho-
mogeneous linear ordinary differential equations.
Chapter 8 deals with oscillatory motion that is central to the description of
acoustics and the effects of inter-particle interaction in many physical systems. In its
most accessible form, oscillatory motion is simple harmonic (s-h). (s-h) motion has
a long and distinguished history of use in modeling physical phenomena. An
harmonic motion—which somewhat more realistically represents the observed
behavior of physical systems—is described next. Detailed analysis of ‘Transient
State’ motion is presented for a point mass for two different oscillatory systems.
These are:
(1) The point mass, m, is tied to the right end of a long, massless coil spring placed
horizontally in the x-direction on top of a long, level, table. The left end of the
coil is fixed to the left end of the long table. The motion of the mass is slowed
by frictional force that is proportional to its momentum m vðtÞ. In its completely
relaxed state, the spring is in equilibrium and the mass is in its equilibrium
position.
(2) Because the differential equations needed for analyzing damped oscillating
pendula are prototypical of those used in studies of electromagnetism, acous-
tics, mechanics, chemical and biological sciences and engineering, we analyze
next a pendulum consisting of a (point-sized) bob of mass m that is tied to the
end of a massless stiff rod of length l. The rod hangs down, in the negative
z-direction, from a hook that has been nailed to the ceiling. The pendulum is set
to oscillate in two-dimensional motion in the x z plane. Air resistance is
approximated as a frictional force proportional to the speed with which the bob
is moving. The ensuing friction slows the oscillatory motion.
In Chap. 9, physics relating to, and the applicable mathematics for the use of,
resistors, inductors, and capacitors are studied. The study includes Kirchhoff’s two
rules that state: ‘The incoming current at any given point equals the outgoing
current at that point’ and ‘The algebraic sum of changes in potential encountered by
charges traveling, in whatever manner, through a closed-loop circuit is zero.’
Considered next is Ohm’s law: namely ‘In a closed-loop circuit that contains a
battery operating at V volt, and a resistor of strength R ohms, current flow is
I amperes: I ¼ VR :’ Problems relating to additions of finite numbers of resistors,
placed in various configurations—some in series and some in parallel formats—are
x Preface
worked out in detail. Also included are several, more involved, analyses relating to
current flows in a variety of infinite networks.
Numerical solutions are analyzed in Chap. 10. Given a first-order linear differ-
ential equation [refer to (10.2)] and its solution, yðx0 Þ; at a point x ¼ x0 ; Runge–
Kutta procedure is used to calculate Yðx0 þ DÞ; an estimate of the exact result
yðx0 þ DÞ: Runge-Kutta estimate is the least accurate when it uses only one step for
the entire move. And indeed, as shown in (10.11), the single-step process does yield
grossly inaccurate results. The two-step process—see (10.12) and (10.13)—im-
proves the results only slightly. But the four-step
effort—see (10.14)–(10.17)—does
1
much better. It reduces the error to about 50 th of that for the one-step process.
Estimates from a ten-step Runge-Kutta process are recorded in table (10.1).
These estimates—being in error only by 100 0:0025 22:17 ¼ 0:0113%—are highly
accurate.
Coupled first-order differential equations (10.18) are treated next.
Together these equations are equivalent to a single second-order differential
equation. Tables (10.2)–(10.7) display numerical results, Xn and Yn , for the
one-step, two-step, and the five-step processes. Table (10.8) contains numerical
results Xn gathered during a twenty-step Runge-Kutta process. At maximum
extension, D ¼ 2, the Runge-Kutta estimate Xn is 26:371190. It differs from the
exact result, 26:371404, by only a tiny amount, 0:000214. The percentage error
involved is 0:000811.
Table (10.9) records numerical results Yn collected during a twenty-step
Runge-Kutta process. At maximum extension, D ¼ 2, the Runge-Kutta estimate Yn
is 11:0170347, It differs from the exact result, 11:0171222, by 0:0000875. The
percentage error involved is 0:000794: It is similar to the corresponding error,
0:000811%, for Xn . The accuracy achieved by the twenty-step Runge-Kutta esti-
mate is quite extraordinary. When very high accuracy is desired, the twenty-step
Runge-Kutta estimate yields results that are worth the effort.
Chapter 11 deals with Frobenius’ work. As stated earlier,
linear (ODEs) with variable coefficients are generally hard to treat. Fortunately,
Frobenius’ method may often be helpful in that regard. To that purpose, analytic
functions, ordinary points, and regular and irregular singular points are described in
this chapter. Frobenius assumes a modified Taylor series solution that is valid in the
neighborhood of an ordinary point. The unknown constants there are determined
through actual use of the Taylor series solution.
Frobenius solution around ordinary point is worked out for differential equations
of type (a)—see (11.5)–(11.30)—and differential equations of type (b)—see (11.32)–
(11.60).
Equations of type (c), (11.62), around regular singular points are treated next–
see (11.63)–(11.75).
Indicial equation is defined—see (11.76)—and equations of category (1), whose
two roots differ by non-integers, of category (2), whose two roots are equal, and of
category (3), whose two roots differ by an integer, are all analyzed [See (11.78)–
(11.183)].
Preface xi
The last part of Chap. 9 deals with Bessel’s equations. Details of the relevant
analyses are provided in (11.184)–(11.242).
Answers to assigned problems are given in Chap. 12.
Fourier transforms and Dirac33: delta function are treated in the appendix which
forms Chap. 13.
Bibliography is presented last.
Unlike a novel, which is often read continuously—and the reading is completed
within a couple of days—this book is likely to be read piecemeal—a chapter or two
a week. At such slow rate of reading, it is often hard to recall the precise form of a
relationship that appeared in the previous chapter. To help relieve this difficulty,
when needed, the most helpful explanation of the issue at hand is repeated briefly
and the most relevant expressions are referred to by their equation numbers.
Throughout the book, for efficient reading, most equations are numbered in seri-
atim. When needed, they can be accessed quickly.
Most of the current knowledge of differential equations is much older than those
of us who study it. The present book owes in motivation to a famous treatise by
Piaggio10: —first published nearly a century ago by G. Bell and Sons, LTD., and
last reprinted in (1940). Piaggio is a great book, but in some important places it
misses, and sometime misprints, relevant detail. Numerous other books11:21: of
varying usefulness are also available. The current text stands apart from these books
in that it is put together with a view to being accessible to all interested readers: for
use both as a textbook and a book for self-study.
Answers to problems suggested for solution are appended in Chap. 12.
Finally, but for the support of my colleague Robert Intemann, this book could
not have been written.
Philadelphia, USA Raza Tahir-Kheli

May 2018
Contents
1 Differential Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 D....................... . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Laws of Addition . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.2 Laws of Multiplication . . . . . . . . . . . . . . . . . . . . . . 2
1.1.3 What is D1 f ðxÞ? . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.4 Index Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Some Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1 Ordinary Differential Equation . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.1 Explicit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.2 Implicit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.3 Linear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.4 Homogeneous . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.5 Inhomogeneous . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.6 Nonlinear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.7 Partial Differential Equation . . . . . . . . . . . . . . . . . . . 8
2.1.8 The Order of an (ODE) . . . . . . . . . . . . . . . . . . . . . . 8
2.1.9 The Degree of an (ODE) . . . . . . . . . . . . . . . . . . . . . 8
2.1.10 Order and Degree: Exercises . . . . . . . . . . . . . . . . . . 8
2.1.11 Characteristic Equation: Ech . . . . . . . . . . . . . . . . . . . 9
2.1.12 Complementary Solution: Scomp . . . . . . . . . . . . . . . . 9
2.1.13 Particular Integral: Ipi . . . . . . . . . . . . . . . . . . . . . . . 10
2.1.14 Indicial Equation . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.1.15 General Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.1.16 Complete Solution . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.1.17 Complete Primitive . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1.18 Singular Solution . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 How Some (ODE) Arise . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
xiii
xiv Contents
3 Constant Coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1 Homogeneous Linear (ODEs) . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1.1 First Order . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1.2 Second Order . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.1.3 Characteristic Equation: Ech . . . . . . . . . . . . . . . . . . . 15
3.1.4 Unequal Roots . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.1.5 Complementary Solution . . . . . . . . . . . . . . . . . . . . . 15
3.1.6 Examples Group I: Unequal Real Roots . . . . . . . . . . 16
3.1.7 Examples Group II: Complex Roots . . . . . . . . . . . . 17
3.1.8 Equations with Complex Roots . . . . . . . . . . . . . . . . 17
3.1.9 Equation with Double Root . . . . . . . . . . . . . . . . . . . 18
3.1.10 n-Equal Roots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.1.11 Ech with Multiple Roots . . . . . . . . . . . . . . . . . . . . . 20
3.1.12 Problems Group I . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2 Linear Dependence and Linear Independence . . . . . . . . . . . . . 21
3.2.1 Wronskian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2.2 Examples Group III . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2.3 Examples Group IV . . . . . . . . . . . . . . . . . . . . . . . . 25
3.3 Method of Undetermined Coefficients . . . . . . . . . . . . . . . . . . . 27
3.3.1 Particular Integral: Ipi . . . . . . . . . . . . . . . . . . . . . . . 28
3.3.2 Examples Group V . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.3.3 Examples Group VI . . . . . . . . . . . . . . . . . . . . . . . . 30
3.3.4 Problems Group II . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.3.5 Examples Group VII . . . . . . . . . . . . . . . . . . . . . . . . 35
3.3.6 Problems Group III . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.3.7 Ipi for BðxÞ ¼ cosðaxÞ ; sinðaxÞ . . . . . . . . . . . . . . . . 37
3.3.8 Examples Group VIII . . . . . . . . . . . . . . . . . . . . . . . 37
3.3.9 Problems Group IV . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.3.10 Ipi for BðxÞ ¼ expða xÞ WðxÞ . . . . . . . . . . . . . . . . . . 40
3.3.11 Examples Group IX . . . . . . . . . . . . . . . . . . . . . . . . 41
3.3.12 Problems Group V . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.4 Simultaneous Linear (ODEs) with Constant Coefficients . . . . . 46
3.4.1 Separable Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.4.2 Problems Group VI . . . . . . . . . . . . . . . . . . . . . . . . . 58
4 Variable Coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.1 Linear (ODE)’s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.1.1 First-Order and First-Degree . . . . . . . . . . . . . . . . . . 59
4.1.2 Integrating Factor . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.1.3 Equation (4.2): Solution . . . . . . . . . . . . . . . . . . . . . 61
4.2 Examples Group I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.2.1 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.2.2 Problems Group I . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Contents xv
4.3 Bernouilli Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

4.3.1 The Bernouilli Suggestion . . . . . . . . . . . . . . . . . . . . 65
4.3.2 Examples Group II . . . . . . . . . . . . . . . . . . . . . . . . . 67
5 Green’s Function Laplace Transforms . . . . . . . . . . . . . . . . . . . . . . 75
5.1 Green’s Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.2 Solving Differential Equations . . . . . . . . . . . . . . . . . . . . . . . . 77
5.2.1 Eigenfunction Expansion . . . . . . . . . . . . . . . . . . . . . 77
5.2.2 Green’s Function Calculated . . . . . . . . . . . . . . . . . . 79
5.2.3 Examples Group I . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.2.4 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.3 Calculation by Approaching Delta Function . . . . . . . . . . . . . . 83
5.3.2 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.3.4 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
5.4 Laplace Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.4.1 Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
5.5 Computation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.6 Heaviside Step Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.6.1 On- and Off Switches . . . . . . . . . . . . . . . . . . . . . . . 96
5.7 Solving Initial Value Problems . . . . . . . . . . . . . . . . . . . . . . . 97
5.8 First-Order Differential Equations . . . . . . . . . . . . . . . . . . . . . . 97
5.8.1 Partial Fraction Decomposition of (5.100) . . . . . . . . 98
5.9 Second-Order Differential Equations . . . . . . . . . . . . . . . . . . . . 108
5.9.1 Solution by Laplace transform . . . . . . . . . . . . . . . . . 108
5.10 Need for Convolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.11 Convolution Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
6 Special Types of Differential Equations . . . . . . . . . . . . . . . . . . . . . . 119
6.1 Clairaut Equation: Description . . . . . . . . . . . . . . . . . . . . . . . . 119
6.1.1 Solving Clairaut Equation . . . . . . . . . . . . . . . . . . . . 120
6.1.3 Singular Solution . . . . . . . . . . . . . . . . . . . . . . . . . . 122
6.1.4 Equations I(B)–I(F) . . . . . . . . . . . . . . . . . . . . . . . . . 124
xvi Contents
6.1.5 Informal Solution . . . . . . . . . . . . . . . . . . . . . . . . . . 124

6.1.6 Formal Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
6.1.7 Problems Group I . . . . . . . . . . . . . . . . . . . . . . . . . . 126
6.2 Lagrange Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
6.2.2 Examples II(A)–II(C) . . . . . . . . . . . . . . . . . . . . . . . 128
6.3 Separable Differential Equations . . . . . . . . . . . . . . . . . . . . . . . 130
6.3.2 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131

6.4 Separable Equations of Form dy ¼ U yx . . . . . . . . . ........ 132
dx
Solution dydx ¼ U x . . . . . . . .. . . . . . . . .
y
6.4.1 ........ 132
Examples Group IV: dy dx ¼ U x Equations
y
6.4.2 ........ 133
6.4.3 Solution . . . . . . . . . . . . . . . . . . . . . . . . . ........ 133
6.4.4 Problems Group IV . . . . . . . . . . . . . . . . . ........ 134

Equations Reducible to dy dx ¼ U x
y
6.5 ............. ........ 135
6.5.1 Examples Group V: Equations (I)–(IV) . . ........ 135
a 1 x þ b 1 y þ c1
6.5.2 Equations dy dx ¼ a2 x þ b2 y þ c2 . . . . . . . . . . . . . . . . . . . . 138
6.5.3 Problems Group V . . . . . . . . . . . . . . . . . . . . . . . . . 141
6.6 Exact Differential and Exact Differential Equation . . . . . . . . . . 142
6.6.1 Exact Differential . . . . . . . . . . . . . . . . . . . . . . . . . . 142
6.6.2 Exact Differential Equation . . . . . . . . . . . . . . . . . . . 143
6.6.3 Requirement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
6.6.4 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
6.6.6 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
6.6.7 Problems Group VI . . . . . . . . . . . . . . . . . . . . . . . . . 147
6.7 Inexact Differential Equation Integrating Factor . . . . . . . . . . . 147
6.7.1 Integrating Factor Dependent only on x . . . . . . . . . . 148
6.7.2 Integrating Factor Dependent only on y . . . . . . . . . . 149
6.7.4 Problems Group VII . . . . . . . . . . . . . . . . . . . . . . . . 153
6.8 Riccati Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
6.8.1 Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
6.8.3 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
6.8.4 Problems Group VIII . . . . . . . . . . . . . . . . . . . . . . . 158
6.9 Euler Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
6.9.1 Examples Group IX . . . . . . . . . . . . . . . . . . . . . . . . 160
6.9.2 Euler Equation: An Extension . . . . . . . . . . . . . . . . . 163
6.9.3 Examples Group X . . . . . . . . . . . . . . . . . . . . . . . . . 165
Contents xvii
6.9.4 Problems Group IX . . . . . . . . . . . . . . . . . . . . . . . . . 167

6.10 Factorable Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
6.10.1 Examples Group XI . . . . . . . . . . . . . . . . . . . . . . . . 169
6.10.2 Problems Group VII . . . . . . . . . . . . . . . . . . . . . . . . 174
6.11 Additional Riccati Equations . . . . . . . . . . . . . . . . . . . . . . . . . 174
6.11.1 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
6.11.2 Additional Examples Group VIII . . . . . . . . . . . . . . . 176
6.11.3 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
6.11.4 Additional Problems Group VIII . . . . . . . . . . . . . . . 179
6.12 Additional Euler Equations . . . . . . . . . . . . . . . . . . . . . . . . . . 179
6.12.1 Additional Examples IX: Euler Equation . . . . . . . . . 181
6.12.2 Solution Additional Examples IX-(A) . . . . . . . . . . . 182
6.12.3 Solution Additional Examples IX-(B) . . . . . . . . . . . . 182
6.12.4 Additional Extended Euler Equations . . . . . . . . . . . . 184
6.12.5 Additional Examples Group X: Extended Euler
Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
6.12.6 Additional Problems Group IX . . . . . . . . . . . . . . . . 188
6.13 Additional Factorable Equations . . . . . . . . . . . . . . . . . . . . . . . 189
6.13.1 Examples Group XI: Additional Factorable
Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
7 Special Situations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
7.1 Ordinary Differential Equations . . . . . . . . . . . . . . . . . . . . . . . 195
7.2 Simple Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
7.2.1 Problems Group I . . . . . . . . . . . . . . . . . . . . . . . . . . 198
7.2.2 Equations (a) and (b) . . . . . . . . . . . . . . . . . . . . . . . 198
7.2.3 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
7.2.4 Equations (c) and (d) . . . . . . . . . . . . . . . . . . . . . . . 200
7.3 Order Reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
7.3.1 Linear Independence of u1 and u2 . . . . . . . . . . . . . . 203
7.4 Reduce Order from Second to First . . . . . . . . . . . . . . . . . . . . 203
7.4.1 Examples (I)–(VII) . . . . . . . . . . . . . . . . . . . . . . . . . 205
7.4.2 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
7.4.4 Particular Integral Ipi ðxÞ . . . . . . . . . . . . . . . . . . . . . 207
7.4.5 Calculation of Ipi ðxÞ Variation of Parameters . . . . . . 207
7.4.6 Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
7.4.7 Examples: Ipi ðxÞ ð1Þ ! Ipi ðxÞ ð4Þ . . . . . . . . . . . 209
7.4.8 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
7.4.9 Examples: Ipi ðxÞ ð5Þ ! Ipi ðxÞ ð10Þ . . . . . . . . . . 212
7.4.10 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
7.4.11 Problems Group IV . . . . . . . . . . . . . . . . . . . . . . . . . 215
xviii Contents
7.5 Other Special Situations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216

7.6 Transformation Using M1 ðxÞ (7.63) . . . . . . . . . . . . . . . . . . . . 216
7.6.2 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
7.6.3 Problems Group V . . . . . . . . . . . . . . . . . . . . . . . . . 222
7.7 Transformation Using M0 ðxÞ . . . . . . . . . . . . . . . . . . . . . . . . . 222
7.8 Examples Group IX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
7.8.1 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
8 Oscillatory Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
8.1 Periodic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
8.1.1 Harmonic Oscillation . . . . . . . . . . . . . . . . . . . . . . . 229
8.1.2 Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
8.1.3 Energy Conservation and Equation of Motion . . . . . 231
8.1.4 Anharmonic Damped Motion . . . . . . . . . . . . . . . . . 232
8.1.5 Over-Damped Anharmonic Motion . . . . . . . . . . . . . 233
8.1.6 Both r1 and r2 Positive Mass Stays on Original
Side . . . . . . . . . . . . . . . . . . . . . . . . .
........ . . . 233
r2
8.1.7 Over-Damped Anharmonic Motion r1 Negative
But > −1 Mass Stays on Original Side . . . . . . . . . . . 235
8.1.8 Over-Damped Anharmonic Motion Equation (8.31)
Not-Satisfied Mass Crosses over . . . . . . . . . . . . . . . 236
8.1.9 Critically Damped Anharmonic Motion . . . . . . . . . . 236
8.1.10 An Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
8.1.11 Under-Damped Spring . . . . . . . . . . . . . . . . . . . . . . 239
8.2 Oscillating Pendulum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
8.2.1 Over-Damped Oscillating Pendulum . . . . . . . . . . . . 241
8.2.2 Angle ht . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
8.2.3 Angle ht and Its Extrema . . . . . . . . . . . . . . . . . . . . 242
8.2.4 Bob Stays on Original Side Curves A–C . . . . . . . . . 242
8.2.5 Extremum in ht : . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
8.2.6 Critically Damped Pendulum . . . . . . . . . . . . . . . . . . 247
8.2.7 Under-Damped Motion . . . . . . . . . . . . . . . . . . . . . . 250
8.2.8 Steady-State Motion . . . . . . . . . . . . . . . . . . . . . . . . 252
8.2.9 Sinusoidal External Force: Steady State . . . . . . . . . . 253
9 Resistors, Inductors, Capacitors . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
9.1 Electric Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
9.1.1 Kirchhoff’s Rules . . . . . . . . . . . . . . . . . . . . . . . . . . 257
9.1.2 Ohm’s Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
9.1.3 Addition of Resistors . . . . . . . . . . . . . . . . . . . . . . . 264
9.1.4 Solution (A) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
9.1.5 Problem (B) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
Contents xix
9.1.6 Solution (B) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269

9.1.7 Problem (C) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270
9.1.8 Solution (C) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
9.2 Infinite Networks of Resistors . . . . . . . . . . . . . . . . . . . . . . . . 271
9.2.1 Current Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
9.2.2 Current I1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
9.2.3 Current Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
9.2.4 Effective Total Resistance . . . . . . . . . . . . . . . . . . . . 276
9.2.5 The Result for ReffectCc . . . . . . . . . . . . . . . . . . . . . 276
9.2.6 ReffectCc as a Function of R1 and R2 . . . . . . . . . . . . 277
9.2.7 Table 1 : Electric Circuit Elements . . . . . . . . . . . . . 278
9.3 Inductors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
9.3.1 Two Inductors in Series . . . . . . . . . . . . . . . . . . . . . 279
9.3.2 Two Inductors in Parallel . . . . . . . . . . . . . . . . . . . . 279
9.3.3 Infinite Network of Inductors . . . . . . . . . . . . . . . . . . 280
9.3.4 Capacitors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280
9.3.5 Charging a Capacitor . . . . . . . . . . . . . . . . . . . . . . . 280
9.3.6 Two Capacitors in Parallel . . . . . . . . . . . . . . . . . . . 281
9.3.7 Two Capacitors in Series . . . . . . . . . . . . . . . . . . . . 282
9.3.8 Infinite Network of Capacitors . . . . . . . . . . . . . . . . . 282
9.3.9 Comment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
9.4 R-C Series-Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
9.4.1 Examples Group I . . . . . . . . . . . . . . . . . . . . . . . . . . 284
9.4.2 R-L Series Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . 285
9.4.4 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
9.4.5 L-C Series-Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . 286
9.5 L-R-C Series Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
9.5.1 Examples Group IV . . . . . . . . . . . . . . . . . . . . . . . . 288
9.5.2 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
9.5.3 Over-Damped Series Circuit . . . . . . . . . . . . . . . . . . 291
9.5.4 Current Flow Across Capacitor . . . . . . . . . . . . . . . . 292
9.5.5 Critically Damped Series Circuit . . . . . . . . . . . . . . . 294
9.5.6 Examples Group V . . . . . . . . . . . . . . . . . . . . . . . . . 295
10 Numerical Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
10.1 Single First-Order Differential Equations . . . . . . . . . . . . . . . . 304
10.1.1 Runge–Kutta Steps . . . . . . . . . . . . . . . . . . . . . . . . . 304
10.1.2 Runge–Kutta Solution . . . . . . . . . . . . . . . . . . . . . . . 305
10.2 Coupled Differential Equations First Order . . . . . . . . . . . . . . . 309
xx Contents
10.2.1 Runge–Kutta Steps . . . . . . . . . . . . . . . . . . . . . . . . . 311

10.2.2 Numerical Solution . . . . . . . . . . . . . . . . . . . . . . . . . 311
11 Frobenius Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
11.1 Normalized Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
11.1.1 An Analytic Function . . . . . . . . . . . . . . . . . . . . . . . 317
11.1.2 Ordinary Point . . . . . . . . . . . . . . . . . . . . . . . . . . . . 318
11.1.3 Regular Singular Point . . . . . . . . . . . . . . . . . . . . . . 318
11.1.4 Irregular Singular Point . . . . . . . . . . . . . . . . . . . . . . 318
11.1.5 Solution Around Ordinary Point . . . . . . . . . . . . . . . 318
11.1.6 Equations of Type (a) . . . . . . . . . . . . . . . . . . . . . . . 318
11.1.7 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
11.1.8 Examples Group I . . . . . . . . . . . . . . . . . . . . . . . . . . 322
11.1.9 Problems Group I . . . . . . . . . . . . . . . . . . . . . . . . . . 324
11.1.10 Equations of Type (b) . . . . . . . . . . . . . . . . . . . . . . . 324
11.1.11 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324
11.2 Frobenious Solution Around Regular Singular Point . . . . . . . . 332
11.2.1 Equations of Type (c) . . . . . . . . . . . . . . . . . . . . . . . 332
11.2.2 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332
11.3 Indicial Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
11.4 Indicial Equation Roots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
11.4.1 m1 and m2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
11.4.3 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
11.5 Examples Group IV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344
11.5.1 Solution: (11.99) . . . . . . . . . . . . . . . . . . . . . . . . . . . 345
11.5.2 General Solution: (11.99)-A . . . . . . . . . . . . . . . . . . 346
11.6 First Solution of (11.99)-(B)!(E) . . . . . . . . . . . . . . . . . . . . . 347
11.7 Methodology For Second Solution . . . . . . . . . . . . . . . . . . . . . 349
11.7.1 Piaggio-Like Solution . . . . . . . . . . . . . . . . . . . . . . . 349
11.7.2 Method of Order Reduction . . . . . . . . . . . . . . . . . . . 351
11.7.3 Complete Solution of (11.99)-(A) . . . . . . . . . . . . . . 353
11.7.4 General Solution of (11.99)-(G) . . . . . . . . . . . . . . . . 354
11.7.5 First Solution of (11.99)-(G) . . . . . . . . . . . . . . . . . . 356
11.7.6 Second Solution of (11.99)-(G) . . . . . . . . . . . . . . . . 357
11.7.7 Piaggio-Like Second Solution . . . . . . . . . . . . . . . . . 357
11.8 Examples Group V . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
11.8.1 General Solution of (11.151)-(4) . . . . . . . . . . . . . . . 358
11.8.2 First Solution of (11.151)-(1)–(5) . . . . . . . . . . . . . . . 360
11.9 Equation (11.151)-(4) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362
Contents xxi
11.9.1 Second Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . 362

11.9.2 Piaggio’s Solution . . . . . . . . . . . . . . . . . . . . . . . . . . 362
11.10 Solution by Method of Order Reduction . . . . . . . . . . . . . . . . . 364
11.11 Bessel’s Equation of Order Zero . . . . . . . . . . . . . . . . . . . . . . 366
11.11.2 First Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369
11.11.3 Piaggio-Like Second Solution . . . . . . . . . . . . . . . . . 369
11.11.4 Solution by Method of Order Reduction . . . . . . . . . 371
11.12 Bessel’s Equation of Order nb . . . . . . . . . . . . . . . . . . . . . . . . 373
11.12.1 Indicial Equation . . . . . . . . . . . . . . . . . . . . . . . . . . 374
11.12.3 Two Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375
11.12.4 First Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375
11.12.5 Second Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
11.13 Bessel’s Indicial Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . 378
11.13.1 Bessel’s Equation of Order Unity . . . . . . . . . . . . . . 379
12 Answer to Assigned Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383
12.1 Problems Group I, 3-chapt . . . . . . . . . . . . . . . . . . . . . . . . . . . 383
12.2 Problems Group II, 3-chapt . . . . . . . . . . . . . . . . . . . . . . . . . . 383
12.3 Problems Group III, 3-chapt . . . . . . . . . . . . . . . . . . . . . . . . . 384
12.4 Problems Group IV, 3-chapt . . . . . . . . . . . . . . . . . . . . . . . . . 385
12.5 Problems Group V, 3-chapt . . . . . . . . . . . . . . . . . . . . . . . . . . 386
12.6 Problems Group VI, 3-chapt . . . . . . . . . . . . . . . . . . . . . . . . . 386
12.14 Problems Group VI, 6-chapt . . . . . . . . . . . . . . . . . . . . . . . . . 391
12.15 Problems Group VII, 6-chapt . . . . . . . . . . . . . . . . . . . . . . . . . 392
12.16 Problems Group VIII, 6-chapt . . . . . . . . . . . . . . . . . . . . . . . . 392
12.17 Problems Group IX, 6-chapt . . . . . . . . . . . . . . . . . . . . . . . . . 392
xxii Contents
12.23 Problems Group I, 11-chapt . . . . . . . . . . . . . . . . . . . . . . . . . . 396

12.24 Problems Group II, 11-chapt . . . . . . . . . . . . . . . . . . . . . . . . . 398
13 Answer to Additional Assigned Problems . . . . . . . . . . . . . . . . . . . . 403
13.1 Fourier Transforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
13.2 Examples Set (I) and Solution . . . . . . . . . . . . . . . . . . . . . . . . 403
13.3 Solution to Examples Set (I) Fourier transforms . . . . . . . . . . . 404
13.4 Dirac’s Delta Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407
Chapter 1
Differential Operator
1.1 D
All differential equations make use of the differential operator D. A convenient

notation for it is the following:
d
D0 = 1 ; D1 = D = ;
dx
dν
Dν = , ν = 2, 3, 4, . . . (1.1)
dx ν
The operator D follows some, but not all, laws of algebra.
1.1.1 Laws of Addition
Assume ν, ν1 , and ν2 are positive integers. Using the rules of calculus, we can check
that D obeys the commutative law of addition,
dν1 f (x) dν2 f (x) dν2 f (x) dν1 f (x)

+ = +
d x ν1 d x ν2 d x ν2 d x ν1
= (D + D ) f (x) = (D + D ν1 ) f (x) ,
ν1 ν2 ν2
(1.2)
as well as the distributive law of addition–multiplication.

ν1
dν d d ν2 d ν d ν1 d ν d ν2
+ f (x) = f (x) + f (x)
d x ν d x ν1 d x ν2 d x ν d x ν1 d x ν d x ν2
= D ν (D ν1 + D ν2 ) f (x) = D ν D ν1 f (x) + D ν D ν2 f (x) . (1.3)
Also the associative law of addition is obeyed.
© Springer Nature Switzerland AG 2018 1

R. Tahir-Kheli, Ordinary Differential Equations,
https://doi.org/10.1007/978-3-319-76406-1_1
2 1 Differential Operator
ν1
dν d ν1 d ν2 dν d d ν2
+ + f (x) = + + f (x)
d xν d x ν1 d x ν2 d xν d x ν1 d x ν2
ν
= (D ν + D ν1 ) + D ν2 f (x) = D + (D ν1 + D ν2 ) f (x). (1.4)
1.1.2 Laws of Multiplication
Similarly, both the associative law,

ν1 ν
dν d d ν2 d d ν1 d ν2
. . f (x) = . . f (x)
d xν d x ν1 d x ν2 d x ν d x ν1 d x ν2
= D ν . (D ν1 . D ν2 ) f (x) = (D ν . D ν1 ) . D ν2 f (x) , (1.5)
and the commutative law,

ν1 ν2
d d ν2 d d ν1
. f (x) = . f (x)
d x ν1 d x ν2 d x ν2 d x ν1
= D ν1 D ν2 f (x) = D ν2 D ν1 f (x) , (1.6)
of multiplication are obeyed. In other words, for ν1 and ν2 positive integers, the
operators D ν1 and D ν2 commute. So far so good! But problems begin to arise if
inverse powers are involved.
1.1.3 What is D−1 f (x)?
First thing to note here is that D represents the process of differentiation with respect
to the variable x. Therefore, quite reasonably, the inverse process—meaning integra-
tion with respect to x—should be represented by D −1 . Simple analysis helps illustrate
this point. For instance, straightforward integration of the differential equation,
D y(x) = 2 x ,
leads to its solution

2 x . dx = x 2 + σ0

dy(x)
= D y(x) . dx = . dx = dy(x) = y(x) (1.7)
dx
1.1 D 3
where σ0 is a constant. If we assume D −1 represents integration with respect to x,

we would also get the same result.

y(x) = D −1 . [D y(x)] = D −1 . (2 x) = (2 x) . dx = x 2 + σ0 .
Another way of looking at this is to write
D −1 . D y = D −1+1 y = D 0 y(x) = y(x) . (1.8)
This suggests two things. First, the operator D −1 may be identified with the operation
of integration. Second, that D −1 . D = D −1+1 may be set equivalent to D 0 = 1.
This looks very promising. But wait a minute !
In view of (1.8), for integral ν that is >μ, the following should hold:
D −ν . D ν x μ = D −ν . 0
= x μ . (1.9)
On the other hand, D ν . D −ν x μ can be written as

x μ+ν
D ν . D −ν x μ = D ν . + O x ν−1 = xμ .
(μ + 1)(μ + 2) . . . (μ + ν)
(1.10)
The two results (1.9) and (1.10) are different. Indeed, whenever ν > μ,
D −ν . D ν x μ = D ν . D −ν x μ . (1.11)
1.1.4 Index Law
In (1.7)–(1.11), we have made use of the index law that asserts:

ν1 ν2
dν1 +ν2 d d
f (x) = . f (x)
d x ν1 +ν2 d x ν1 d x ν2
= D ν1 +ν2 f (x) = D ν1 . D ν2 f (x). (1.12)
Clearly, this law holds if both ν1 and ν2 are positive integers. But above we have used
this law even when one of these indices was not positive? Let us try a very simple
exercise to check the adequacy of the index law when that is the case.
4 1 Differential Operator

−1 x3
D x = D .D
3−1 2 3
x =D .
2 3
+ σ0 = 2 , (1.13)
3
but
D −1+3 x 2 = D −1 . [D 3 x 2 ] = D −1 . [0] = σ1 . (1.14)
Thus, an additional instruction would be needed to set σ1 = 2 . To sum up:
When n is a positive integer, D n may be treated as an ordinary algebraic

quantity. In contrast, negative powers of D are not simple algebraic quantities
and must be treated with care.
Chapter 2
Some Definitions
A wealth of terminology is used in the literature. Unfortunately, some sounds similar

even when it has distinctly different meaning. Noted below are a few definitions that
are used in this manuscript.
2.1 Ordinary Differential Equation
A function involving a dependent variable, say u, only a single independent variable,

say x, and at least one derivative of u with respect to x is an ‘ordinary differential
equation’ (ODE).
(ODE)s are used in problems in biology, chemistry, engineering, physics, other
science disciplines, as well as computational mathematics.
2.1.1 Explicit
An (ODE) may be expressed in the following explicit form
Fx pl (x, u, u (1) , . . . , u (ν−1) ) = u (ν) . (2.1)
Here, u (ν) is the νth differential of the dependent variable u with respect to the
independent variable x, i.e.,
dν u
u (ν) = ≡ D ν u, (2.2)
dx ν
and Fx pl is a continuous function of the variables x, u, u (1) , . . . , u (ν−1) , where the

independent variable x ranges within some specified interval I ≡ X i < x < X f .

https://doi.org/10.1007/978-3-319-76406-1_2
6 2 Some Definitions
Such an equation is called an explicit ordinary differential equation. Knowledge of

a function u(x) defined within the same interval I that satisfies the same differential
equation, i.e.,

Fx pl x, u(x), u (1) (x), . . . , u (ν−1) (x) = u (ν) (x), (2.3)
is equivalent to having an explicit solution of the explicit ordinary differential equa-

tion.
2.1.2 Implicit
When a given (ODE) cannot be expressed in the explicit from, it is said to be an

implicit ordinary differential equation represented as
Fimpl (x, u, u (1) , . . . , u (ν−1) , u (ν) ) = 0. (2.4)
Generally, explicit ordinary differential equations are easier to treat than implicit
ordinary differential equations. For simplicity, therefore, we shall work mostly with
the explicit ordinary differential equations.
2.1.3 Linear
A differential equation that is of the first degree in its dependent variable as well as
all its derivatives and is of the form
ν

ai (x)D i u = B(x) , (2.5)
i=0
where B(x) and {ai (x)}, i = 0, 1, . . . , ν are all known functions of x, is said to be a
linear ordinary differential equation.
2.1.4 Homogeneous
Whenever the function B(x) is missing from the linear (ODE), the latter is referred
to as homogeneous linear ordinary differential equation. That is
ν

ai (x)D i u = 0. (2.6)
i=0
2.1 Ordinary Differential Equation 7
[Unfortunately, this terminology is somewhat confusing because to many readers the

normal usage of the term homogeneous function of degree-n implies the following
equivalence:
f (λx, λy, λz) ≡ λn f (x, y, z). (2.7)
Later in the book, there will be occasion to work with differential equations of the
type
F1 (λx, λy) dx + F2 (λx, λy) dy = λn [F1 (x, y) dx + F2 (x, y) dy , (2.8)
where both F1 (x, y) and F2 (x, y) are homogeneous functions and both are of the
same degree.]
2.1.5 Inhomogeneous
A linear (ODE) of the form (2.5), where B(x) is not equal to zero, is called an
inhomogeneous linear ordinary differential equation.
2.1.6 Nonlinear
When the (ODE) cannot be expressed in linear form represented in (2.5) or (2.6), it is
said to be nonlinear (ODE). For instance, the well-known simple pendulum equation,
where l represents the length, m the mass, and g the acceleration due to gravity—all
of which are assumed to be constants—
d2 {l tan θ}
m + m g sin θ = 0 (2.9)
dt 2
is nonlinear when the angle θ is not very small compared to a radian. Here, θ is
the angle that the massless pendulum rod of length l makes with the vertical, m is
the mass hanging atthe bottom of the rod, t refers to the time, g is the acceleration
due to gravity, and gl is the angular velocity so that one complete cycle—which
is equivalent to angular rotation 2π—is completed in time √
2π
g . It does, however,
l
become linear for very small θ when both tan θ and sin θ tend to θ.
Just as implicit (ODE) is harder to solve than explicit (ODE), solving nonlinear
(ODE) requires more effort than solving linear (ODE). Furthermore, simple treat-
ment of nonlinear (ODE) cannot be guaranteed to succeed. Fortunately, many of the
physical problems of interest—at least in the first approximation—can be expressed
in terms of explicit ordinary differential equations that are linear. Therefore, much
of our attention will be focused on linear explicit ordinary differential equations.
2.1.7 Partial Differential Equation
An (ODE) has only one independent variable. But physical situations of interest
sometime involve more than one physical property. As a result, the functions repre-
senting them may sometime involve two or more independent variables. And such
dependence is often best expressed by partial differential equations that contain par-
tial derivatives with respect to more than one independent variable.
2.1.8 The Order of an (ODE)
The order of an (ODE) is ν if the highest derivative in the equation is of order ν. For
instance, this is the case in (2.1), (2.3), and (2.4) because the highest derivative there
is u (ν) .
Similarly, the order of a partial (ODE) is the order of the highest partial derivative
present.
2.1.9 The Degree of an (ODE)
Unlike the order of an (ODE), determining its degree requires some care. It is neces-
sary first to arrange the (ODE) in a form where all the differential coefficients in the
differential equation appear with rational and integral powers. When that has been
done, the equation can be written as a polynomial in all the derivatives present. Then,
the power of the derivative of the highest order is the degree of the given (ODE).
2.1.10 Order and Degree: Exercises
Determine the order and degree of the following differential equations.

2
D3u =A(x) D 2 u + B(x) Du + G(x) , (2.10)
2 √1
D u 2 =u + A(x) , (2.11)
2 23 1 2
D u = 1 + D3u 2 , (2.12)
2 2
∂ u ∂ u
=A(x, y) . (2.13)
∂2 y x ∂2 x y
The highest order derivative in (2.10) is D 3 u. Therefore, it is of third-order. Also, the

differentials that appear here have powers that are integral. And, the highest order
differential occurs with a power of 2. Therefore, (ODE) (2.10) is of second degree.
Regarding (ODE) (2.11), the only and therefore the highest order derivative is
D 2 u. As such, this (ODE) is of second order. To determine its degree, one needs
√ first
to rationalize all its derivatives. To this end, raise both sides to the power 2.
√
D 2 u = [(u + A(x)] 2
. (2.14)
The highest-order derivative is clearly of power unity. Therefore, (ODE) (2.11) is of

first degree.
For (2.12), proceed as follows: Raise both sides to power 23 .

1 3 1 3
D 2 u = 1 + (D 3 u) 2 = 1 + 3 D3u + 3 D3u 2 + D3u 2 .
Now transfer 1 + 3 D 3 u to the left-hand side

1
D 2 u − 3 D 3 u − 1 = (D 3 u) 2 3 + D 3 u
and square both sides.

2 2
D2u − 3 D3u − 1 = D3u 3 + D3u . (2.15)
The highest-order differential in (2.15) is D 3 u. Also, the highest power it occurs

3
with—see the right-hand side—is D 3 u . Therefore, the (ODE) (2.11) is of third
order and third degree.
The partial differential equation (2.13) is clearly of second order and first degree.
2.1.11 Characteristic Equation: E ch
When exp(kx) is tried as a solution to an nth order homogeneous linear equation,

the result appears in the form exp(kx) multiplied by an nth degree polynomial in the
variable k. Such polynomial will be referred to as the characteristic equation of the
given nth order homogeneous linear ordinary differential equation.
2.1.12 Complementary Solution: Scomp
Determine the n solutions of the nth degree polynomial in k—that is, contained in the
characteristic equation mentioned above —and notate them k j with j = 1, 2, . . . , n.
Next, exponentiate each one of them. Then, the sum

n
σ j exp k j , (2.16)
j=1
where σ j are arbitrary constants, is called a complementary solution.
2.1.13 Particular Integral: I pi
A solution of an inhomogeneous linear (ODE) that contains no arbitrary constants

will be called its particular integral I pi .
2.1.14 Indicial Equation
The term with the lowest power of x—such as (11.76)—in a Frobenius solution—
such as (11.75)—is termed the indicial equation.
2.1.15 General Solution
A solution that contains arbitrary constants is often called a general solution. General
solution is a bit of a misnomer because the requirement for a solution to be general
is quite relaxed. Of the possible n arbitrary constants in the solution, only one needs
be nonzero. Occasionally in this text, general solution will refer to a solution given
in terms of the variable ν0 that symbolizes roots of the indicial equation.
2.1.16 Complete Solution
Assume the n functions in the complementary solution—namely, σ j exp k j —every

one of which is itself a solution of the nth order linear (ODE) are linearly independent.
Add the complementary solution to the particular integral I pi . The sum of the two
makes up a complete solution. That is,
Complete Solution = I pi + Complementar y Solution . (2.17)
[Note: Linear independence is explained at length in (3.43)–(3.47). Also note

that by making appropriate adjustment of the n arbitrary multiplying constants,
σi , i = 1, . . . , n, derived from the complete solution all possible solutions can be

obtained. Accordingly, complete solution may have many equivalent forms.]
2.1.17 Complete Primitive
Should all the n arbitrary multiplying constants, σi , i = 1, . . . , n, that occur in the

complementary solution stand determined by the use of appropriate boundary con-
stants then the resultant complete solution is called complete primitive.
2.1.18 Singular Solution
Occasionally, there may exist a solution to a given differential equation that cannot be
derived from its complete solution or complete primitive. Such an unusual solution
is often called a singular solution.
2.2 How Some (ODE) Arise
Upon Elimination of Constants
While an appropriate physical motivation is the source of most differential equations,

differential equations may also arise in other ways. In particular, they may result from
elimination of unknown constants in an equation. For instance, upon differentiation
the following equations with one and two constants may be expressed as first- and
second-order differential equations, respectively. In this manner, the equations
u(x) = σ0 exp (k x) ,
v(x) = σ1 sin(a x) + σ2 cos(a x) , (2.18)
would lead to the following differential equations.
Du = k u ;
D 2 v = −a 2 v . (2.19)
Notice that by eliminating a single constant, e.g., σ0 , a first-order (ODE) is derived

while a differential equation of second order is obtained from eliminating two con-
stants such as σ1 and σ2 . Thus, it is usually the case that the elimination of n constants
leads to a differential equation of the nth order. Conversely, solution to a differential
equation of the nth order usually contains n unknown constants.
As an aside each of the surviving known constants k and a in (2.19) can also
be eliminated by further differentiation which raises the order of each of the two
differential equations by one notch.
u D 2 u = (Du)2 ,
v D 3 v = (D 2 v)(Dv) . (2.20)
Chapter 3
Constant Coefficients
Most differential equations that describe vibrational motion belong to a class of

linear ordinary differential equations where the coefficients are known constants.
User-friendly methods for solving both the homogeneous and the inhomogeneous
versions of these equations are described in this chapter.
3.1 Homogeneous Linear (ODEs)
Homogeneous linear ordinary differential equations with constant coefficients are of

the form
ν

ci D i u(x) = 0 (3.1)
i=0
where ci for i = 0, 1, 2, . . . , etc., are known constants. Exact solution of such equa-
tions can often be found in terms of elementary functions when the equations are
first order or second order, and also—with somewhat greater effort—when they are
third order or fourth order.
3.1.1 First Order
The simplest homogeneous linear ordinary differential equation is of the first order.

1
ci D i u(x) = (c0 + c1 D) u(x) = 0 .
i=0

https://doi.org/10.1007/978-3-319-76406-1_3
14 3 Constant Coefficients
Rewrite the above equation

du(x) c0
= − dx (3.2)
u(x) c1
and integrate.

c0
ln u(x) = − x +σ . (3.3)
c1
The solution is

c0
u(x) = σ0 exp − x , (3.4)
c1
where σ0 = exp(σ) is the single unknown arbitrary constant that can be determined
from one boundary condition. For instance, the value of u(x) at x = 0 is equal to σ0 .
Notice that only one integration is needed to obtain a solution of the first-order
homogeneous linear ordinary differential equation with constant coefficients and the
solution has only a single arbitrary constant. Indeed, the number of independent
constants that a complete solution of an nth order homogeneous linear ordinary
differential equation with constant coefficients contains usually is equal to n and
such a solution requires the equivalent of n integrations.
3.1.2 Second Order
The following is a second-order homogeneous linear ordinary differential equation

with constant coefficients.

c0 + c1 D + c2 D 2 u(x) = 0 . (3.5)
Given our experience with a first-order homogeneous linear ordinary differential

equation with constant coefficients—see (3.4)—let us again try
u(x) = exp (k x) (3.6)
as a possible solution. With this choice, (3.5) becomes

c0 + c1 D + c2 D 2 exp (k x) =

exp (k x) c0 + c1 k + c2 k 2 = 0 . (3.7)
3.1 Homogeneous Linear (ODEs) 15
Except for the extraordinary circumstance when k x → −∞, (3.7) requires

c0 + c1 k + c2 k 2 = 0 . (3.8)
This is the characteristic equation. Note, given a differential equation, e.g. (3.5), the
characteristic equation is readily found by replacing D with k.
3.1.3 Characteristic Equation: E ch
The characteristic equation for all second-order homogeneous linear ordinary differ-
ential equation is a quadratic. The two roots of (3.8) are k1 and k2 .

k1 = (2 c0 )−1 −c1 + c12 − 4 c0 c2 , (3.9)

k2 = (2 c0 )−1 −c1 − c12 − 4 c0 c2 . (3.10)
3.1.4 Unequal Roots
Assuming the two roots, whether they be real or imaginary, are unequal, then the
differential equation (3.5) has two solutions of the form
u(x) = u 1 (x) = exp (k1 x) . (3.11)

u(x) = u 2 (x) = exp (k2 x) . (3.12)
3.1.5 Complementary Solution
And the complementary solution is their linear sum. That is
Scomp = σ1 exp (k1 x) + σ2 exp (k2 x) . (3.13)
As usual, σ1 and σ2 represent arbitrary constants.

Superposition Principle
When exp (k1 x) and exp (k2 x) are solutions of a second-order homogeneous linear
ordinary differential equation with constant coefficients, then a linear sum of the two,
namely the complementary solution,
σ1 exp (k1 x) + σ2 exp (k2 x) , (3.14)
is also a solution of the same homogeneous linear ordinary differential equation.

Thus, the superposition principle asserts that the following must be true:

c0 + c1 D + c2 D 2 .[σ1 exp (k1 x) + σ2 exp (k2 x)] = 0.
(3.15)
Knowing that a given set of derivatives of the product σ u(x) is equal to σ times the
same set of derivatives of u(x), and the fact that derivatives of a sum are equal to the
sum of those derivatives, (3.15) can be rewritten as

σ1 c0 + c1 D + c2 D 2 u 1 (x)

+σ2 c0 + c1 D + c2 D 2 u 2 (x) = 0 . (3.16)
According to (3.9)–(3.13), the relationship (3.16) is indeed satisfied. This confirms

the assertion (3.15) and therefore the superposition principle.
3.1.6 Examples Group I: Unequal Real Roots
Complementary solution is worked out for three differential equations for which the
characteristic equation as well as its roots is provided. Note, these are second-order
homogeneous linear ordinary differential equation with constant coefficients, whose
characteristic equation has two real roots, k1 and k2 , that are unequal.

9 9
D 2 + 4D − u = 0 ; E ch = k 2 + 4k − = 0 ,
4 4
1 9
k1 = , k2 = − . (3.17)
2 2
2
2D + 5D + 3 u = 0 ; E ch = 2k 2 + 5k + 3 = 0 ,
3
k1 = − 1 , k2 = − . (3.18)
2
2
2D − 5D + 3 u = 0 ; E ch = 2k 2 − 5k + 3 = 0 ,
3
k1 = 1 , k2 = . (3.19)
2
As described in (3.5) → (3.13), the relevant complementary solution for the differ-
ential equations in (3.17)–(3.19) is the following.
Scomp = σ1 exp(k1 x) + σ2 exp(k2 x) . (3.20)

3.1.7 Examples Group II : Complex Roots
Treated below is a second-order homogeneous linear ordinary differential equation

with constant coefficients whose characteristic equation is a quadratic with complex
roots k1 and k2 .
k1 = r − i m ; k2 = r + i m . (3.21)
According to (3.20), the complementary equation takes the form
Scomp = σ1 exp (r − i m) x + σ2 exp (r + i m)x

= exp (r x) σ1 exp (−i m x) + σ2 exp (i m x)
= exp (r x) [σ3 sin(m x) + σ4 cos(m x)] , (3.22)
where the arbitrary constants σ3 and σ4 represent combinations of the earlier arbitrary
constants in the form
σ3 = −i (σ1 − σ2 ) ,
σ4 = (σ1 + σ2 ) . (3.23)
In deriving the above equation, the following identity was also used
exp (±i θ) = cos(θ) ± i sin(θ) . (3.24)
3.1.8 Equations with Complex Roots
Given below are differential equations along with their characteristic equations and
complex roots in the form of r and m. [Note: See (3.21) for the definition of r and m.]

D 2 + 2D + 5 u(x) = 0 ; k 2 + 2k + 5 = 0 ;
r = −1 , m = 2 . (3.25)

D 2 − 6D + 10 u(x) = 0 ; k 2 − 6k + 10 = 0 ;
r = 3 , m = 1 . (3.26)

5 5
D 2 + 3D + u(x) = 0 ; k 2 + 3k + = 0 ;
2 2
3 1
r = − , m= . (3.27)
2 2
The relevant complementary solution is obtained by introducing these values of

r and m into (3.22).
3.1.9 Equation with Double Root
Consider next a second-order homogeneous linear ordinary differential equation with

constant coefficients whose characteristic equation is a quadratic with a double root,
real or imaginary. For instance, when the differential equation

D 2 − 2 c D + c2 u(x) = (D − c)2 u(x) = 0 (3.28)
is solved by setting u(x) = exp (k x), its E ch , namely
(k − c)2 = 0 , (3.29)
yields identical two roots k = c. Therefore,
u(x) = u 1 (x) = exp(c x) (3.30)
is a double solution that duly satisfies the differential equation (3.28). But it is only
a single distinct solution. [Note: Two solutions are distinct if they are linearly inde-
pendent. Linear independence is formally defined in (3.43)–(3.45).] Hoping that
another distinct solution would also involve exp(c x), let us try
u(x) = u 2 (x) = exp(c x) f (x) (3.31)
as a possible second distinct solution and determine f (x) accordingly. To this end,
insert (3.31) into (3.28). One gets

0 = (D − c)2 u 2 (x) = (D − c)2 exp(c x) f (x)

= (D − c) (D − c) exp(c x) f (x) = (D − c) {exp(c x) D f (x)}

= exp(c x) D 2 f (x) . (3.32)
Thus, all that is needed is to find the solution to
d
D 2 f (x) = {D f (x)} = 0 . (3.33)
dx
This is readily done by integrating (3.33) twice. The first integration leads to

d
D f (x) dx ≡
2
{D f (x)} dx = [0]dx ,
dx
= {D f (x)} + const 1 = const 2 . (3.34)
And the second to
[D f (x)] dx = (const 2 − const 1 ) dx ,

f (x) + const 3 = (const 2 − const 1 ) (x + const 4 ) . (3.35)
By setting (const 2 − const 1 ) = σ1 and (σ1 . const 4 − const 3 ) = σ0 , one gets
f (x) = σ0 + σ1 x . (3.36)
Thus, according to (3.31) and (3.36), the u(x) given below solves the differential
equation (D − c)2 u(x) = 0.
u(x) = exp(c x) f (x) = exp(c x) (σ0 + σ1 x) . (3.37)
This version of u(x)—which is to be dubbed the solution—is quite interesting. In

addition to including a new expression, namely exp(c x) (σ1 x), it contains also all
of the first solution, namely exp(c x) (σ0 ), that was given in (3.30). Additionally,
because u(x) has been derived from two integrations, it has the requisite number
of unknown constants: namely σ0 and σ1 . Therefore, u(x), given in (3.37), is the
complementary solution. Finally, as long as its two parts—namely (σ0 ) exp(c x)
and (σ1 x) exp(c x)—are mutually distinct, meaning they are linearly independent,
u(x) is also the complete solution of the differential equation (3.28). Additionally,
if the two unknown constants, σ0 and σ1 , are determined—say, from two boundary
conditions—this solution would also qualify as the complete primitive. [Note: See the
succeeding subsection for a discussion of when a given set of functions is mutually
distinct, meaning when they are linearly independent. Also see (3.47)–(3.53) where
the linear independence of given two expressions is demonstrated.]
3.1.10 n-Equal Roots
The foregoing can be extended to nth-order homogeneous linear ordinary differential

equation with constant coefficients.
(D − c)n u(x) = 0 .
Clearly, its characteristic equation

(k − c)n = 0
has n equal roots: k = c. Following the same procedure as outlined in deriving (3.37),
the complementary solution is
Scomp (x) = exp(c x)[σ0 + σ1 x + σ2 x 2 + · · · + σn−1 x n−1 ]. (3.38)
And once again it is a sum of distinct terms. As such, it is a complete solution. If the
values of σi − i = 0 , 1 , 2 , . . . , (n − 1) were all known—say, determined from n
different boundary conditions—the Scomp (x), given above, would also qualify as the
complete primitive.
Characteristic equations with multiple roots are analyzed below.
Equation (3.38) refers to the case where E ch has n equal roots. That means, there
is only one distinct root that occurs n different times. What if there were two roots,
both of them occurring multiple times? To study this case, consider the following
differential equation with two roots: c0 and c.
F u(x) = (D − c0 )n 0 (D − c)n u(x) = (D − c)n (D − c0 )n 0 u(x) = 0 .

(3.39)
3.1.11 E ch with Multiple Roots

Two Roots Equal and Occurring Multiple Times
If the two roots c0 and c are equal, the problem (3.39) is similar to that worked out
in (3.38). Then, the result is

u(x) = Scomp (x) = exp(c x) σ0 + σ1 x + σ2 x 2 + · · · + σn 0 +n−1 x n 0 +n−1
(3.40)
Two Roots, Unequal and Occurring Multiple Times

On the other hand, for the more interesting case, should c0 and c be unequal, the
above differential equation would have two distinct roots: namely c0 , that occurs n 0
times, and c, that occurs n times. The characteristic equation then would be
E ch = (k − c0 )n 0 (k − c)n = (k − c)n (k − c0 )n 0 .
Define
(D − c0 )n 0 U0 (x) = 0 ,
(D − c)n U (x) = 0 .
Then,
(D − c0 )n 0 (D − c)n [U (x) + U0 (x)]
= (D − c0 )n 0 (D − c)n U (x) + (D − c0 )n (D − c)n U0 (x)
= 0 + (D − c0 )n 0 (D − c)n U0 (x) = 0 + (D − c)n (D − c0 )n 0 U0 (x)
= 0+0 . (3.41)
Thus, the solution of homogeneous linear ordinary differential equation with constant
coefficients (3.39) is
Scomp (x) = u(x) = U0 (x) + U (x)

= exp (c0 x) σ00 + σ01 x + σ02 x 2 + · · · + σ0(n 0 −1) x n 0 −1

+ exp(c x) σ0 + σ1 x + σ2 x 2 + · · · + σn−1 x n−1 . (3.42)
3.1.12 Problems Group I
Find complementary solution, Scomp (x), to the following ten homogeneous linear
ordinary differential equations with constant coefficients.

D 2 + 2D − 3 u(x) = 0 . (1)

D 2 − 3D − 4 u(x) = 0 . (2)

1 2
D+ u(x) = 0 . (3)
2

1 2
D− u(x) = 0 . (4)
2
(D + 2)3 u(x) = 0 . (5)
(D − 2)3 u(x) = 0 . (6)

D + D + 1 u(x) = 0
2
. (7)

D 2 − D + 1 u(x) = 0 . (8)

D 2 + 2D + 3 u(x) = 0 . (9)

D 2 + 3D + 4 u(x) = 0 . (10)
3.2 Linear Dependence and Linear Independence

Linear Dependence
A given set of functions { f i (x)}, i = 1, 2, ..., n, is said to be ‘linearly dependent’
if and only if, for all values of x in an interval I , there exist n constants, say {σi },
i = 1, 2, ..., n, such that the following relationship—namely (3.43)—holds true.

n
σi f i (x) = 0 . (3.43)
i=1
Clearly, the case when the constants {σi }, i = 1, 2, ..., n, are all vanishing is trivial.
A straightforward use of the relationship (3.43) for determining linear dependence
of the functions f i (x) is quite awkward. It requires knowledge of an appropriate, non-
trivial, set of constants σi . This requirement is circumvented below.
In order to determine whether the given n functions, { f i (x)}, i = 1, 2, ..., n—each
of which is differentiable (n − 1) times—are linearly dependent, one differentiates
(3.43) several times and keeps a record. Every time (3.43) is differentiated, the pro-
cess produces a new differential equation. This way, after
(n − 1) differentiations, there are (n − 1) new differential equations. These along
with the original equation, namely (3.43), make a total of n simultaneous homoge-
neous linear ordinary differential equations with constant coefficients that involve
the n constants {σi }, i = 1, 2, ..., n. It is helpful to display these equations.

n
σi f i (x) = 0 ; (1)
i=1

n
σi . D f i (x) = 0 ; (2)
i=1
..... . .....

n
σi . D n−1 f i (x) = 0 ; (n) (3.44)
i=1
For the given n functions, { f i (x)}, i = 1, 2, ..., n—each of which is differentiable

(n − 1) times—to be linearly dependent, it is required that (3.44) be satisfied by a
non-trivial choice of the constant σi . To ensure that it is so, proceed as follows.
3.2.1 Wronskian
A well-known theorem of algebra states that n simultaneous homogeneous linear

equations involving n constants—such as σi ’s, i = 1, 2, ..., n, in (3.44)—have a
non-trivial solution if and only if the following relationship—that is, (3.45) given
below—holds true.
3.2 Linear Dependence and Linear Independence 23

f 1 (x) f 2 (x), ..., f n (x)

D f 1 (x) D f 2 (x), ..., D f n (x)

W (x) ≡

............. ....................
D n−1 f 1 (x) D n−1 f 2 (x), ..., D n−1 f n (x)
=0. (3.45)
In other words:
For all values of x within an interval I , the given n functions—{ f i (x)}, i =
1, 2, ..., n—each of which is differentiable at least (n − 1) times, are linearly depen-
dent if and only if their Wronskian, W (x), is vanishing.
Linear Independence
On the other hand, linear independence, being the opposite of linear dependence,
obtains only if a relationship like (3.43) never holds true—except, of course, for
the trivial case when all the constants σi are zero. Indeed, a given set of functions
{ f i (x)}, i = 1, 2, ..., n, is said to be linearly independent if and only if, for all values
of x in an interval I , there exist n non-trivial constants, say {σi }, i = 1, 2, ..., n, such
that relationship (3.46) obtains.

n
σi f i (x) = 0 . (3.46)
i=1
Employing the same argument that led from (3.43) to (3.45), one concludes that
functions { f i (x)}, i = 1, 2, ..., n, are linearly independent if and only if the following
relationship—namely (3.47)—holds true.

f 1 (x) f 2 (x), ..., f n (x)

D f 1 (x) D f (x), ..., D f (x)
W (x) = 2 n = 0 . (3.47)
............. ....................
n−1
D f 1 (x) D n−1 f 2 (x), ..., D n−1 f n (x)
Thus to reiterate: For all values of x within an interval I , functions { f i (x)},

i = 1, 2, ..., n, each of which is differentiable at least (n − 1) times, are linearly
independent if and only if their Wronskian, W (x), is non-vanishing.
3.2.2 Examples Group III
(A): Work out the requirement that must be satisfied for a given pair of functions,
f 1 (x) and f 2 (x), each of which is differentiable at least once, to be linearly dependent.
Solution
According to (3.45), two such functions f 1 (x) and f 2 (x) are linearly dependent if
and only if they obey the relationship

f (x) f 2 (x)
W (x) = 1 =0. (3.48)
D f 1 (x) D f 2 (x)
Rewriting (3.48),
D f 2 (x) D f 1 (x)
− = 0 , (3.49)
f 2 (x) f 1 (x)
and integrating,
log[ f 2 (x)] = log[ f 1 (x)] + constant , (3.50)
gives
f 2 (x) = C0 f 1 (x) . (3.51)
Simply expressed, two functions that are proportional are linearly dependent.
(B): Given u(x) in (3.37) is complementary solution of the homogeneous linear
ordinary differential equation with constant coefficients (3.28) that has equal roots,
show that u(x) is also its complete solution Scs (x).
In order for (3.37) to be a complete solution of the homogeneous linear ordi-
nary differential equation with constant coefficients (3.28), the given two functions
f 1 (x) = exp(c x) and f 2 (x) = x exp(c x) have to be linearly independent. This will
be the case, according to (3.47), if the following holds true.
D f 2 (x) D f 1 (x)
− = 0 . (3.52)
f 2 (x) f 1 (x)
Equation (3.52) translates into

exp(c x) + c x exp(c x) c exp(c x)
− = 0 (3.53)
x exp(c x) exp(c x)
which reduces to the inequality
1
= 0 . (3.54)
x
Clearly, as long as x is non-infinite this inequality holds. Therefore, the given two
functions are linearly independent, and as a result (3.37) is a complete solution of the
homogeneous linear ordinary differential equation with constant coefficients (3.28).
3.2 Linear Dependence and Linear Independence 25
3.2.3 Examples Group IV
Using the methods described in the foregoing, work out complementary solution to
the set of twelve homogeneous linear ordinary differential equations with constant
coefficients given in (3.55) below.

D 2 + 3D + 1 u(x) =0 . (1)

D 2 + 3D − 1 u(x) =0 . (2)

D 2 + 3D − 3 u(x) =0 . (3)

9
D 2 + 3D + u(x) =0 . (4)
2

D 2 + 4D + 6 u(x) =0 . (5)

1
D2 + D + u(x) =0 . (6)
2

3 2
D+ u(x) =0 . (7)
2

3 2
D− u(x) =0 . (8)
2
(D + 2)2 u(x) =0 . (9)
(D + 1)3 u(x) =0 . (10)

D (D 2 − 4)2 u(x) =0 . (11)
2 2
D (D − 4) u(x) =0 . (12) (3.55)
Solutions to Examples Group IV, (3.55)

In the following, complementary solution is worked out for all of the twelve homo-
geneous linear ordinary differential equations with constant coefficients given above
in (3.55). This is done by determining their characteristic equation and its roots.
Remember, as always, the characteristic equation is determined by replacing D by k.
√
3 5
(1) E ch : k + 3k + 1 = 0 ; k1,2 = − ±
2
;
√ 2 2
√
3 5 5
(1) Scomp (x) = exp − x σ1 exp x + σ2 exp − x .
2 2 2
√
3 13
(2) E ch : k + 3k − 1 = 0 ; k1,2 = − ±
2
;
2 2
√ √
3 13 13
2 2 2
√
3 21
(3) E ch : k + 3k − 3 = 0 ; k1,2 = − ±
2
;
√ 2 2
√
3 21 21
2 2 2
9 3 3
(4) E ch : k 2 + 3k + = 0 ; k1,2 = − ± i ;
2
2 2

3 3 3
(4) Scomp (x) = exp − x σ1 sin x + σ2 cos x .
2 2 2
√
(5) E ch : k 2 + 4k + 6 = 0 ; k1,2 = − 2 ± i 2 ;
√ √
(5) Scomp (x) = exp (−2x) σ1 sin x 2 + σ2 cos x 2 .
1 1 1
(6) E ch : k 2 + k + = 0 ; k1,2 = − ± i ;
x 2 x 2 2
x
(6) Scomp (x) = exp − σ1 sin + σ2 cos .
2 2 2
2
3 3
(7) E ch : k + = 0 ; k1,2 = − ;
2 2

3
(7) Scomp (x) = exp − x (σ0 + σ1 x) .
2
2
3 3
(8) E ch : k − = 0 ; k1,2 = ;
2 2

3
(8) Scomp (x) = exp x (σ0 + σ1 x) .
2
(9) E ch : (k + 2)2 = 0 ; k1,2 = − 2 ;
(9) Scomp (x) = exp (−2x) (σ0 + σ1 x) .
(10) E ch : (k + 1)3 = 0 ; k1,2,3 = − 1 ;

(10) Scomp (x) = exp (−x) σ0 + σ1 x + σ2 x 2 .
2
(11) E ch : k k 2 − 4 = 0 ; k1 = 0 , k2,3 = ± 2 ;
(11) Scomp (x) = σ0 + exp (2 x) (σ1 + σ2 x) + exp (−2 x) (σ3 + σ4 x) .

(12) E ch : k 2 k 2 − 4 = 0 ; k1,2 = 0 , k3,4 = ± 2 ;
(12) Scomp (x) = σ0 (1 + σ1 x) + σ2 exp (2 x) + σ3 exp (−2 x) . (3.56)
3.3 Method of Undetermined Coefficients 27
3.3 Method of Undetermined Coefficients
Treatment of Inhomogeneous Linear Ordinary Differential Equation with

Constant Coefficients
Consider
ν

c j D j u(x) = B(x) .
j=0
If B(x) is nonzero, then this equation is an inhomogeneous linear ordinary differential

equation with known constant coefficients, c j .
Assume that through some fantastic luck, one has been able to guess the particular
integral, I pi , that solves the above equation. That is
ν

c j D j I pi = B(x) . (3.57)
j=0
Clearly, by using the procedure outlined in the foregoing subsections, one can also
work out a complementary solution for the homogeneous part of the above inho-
mogeneous linear ordinary differential equation with constant coefficients. In other
words, one can determine Scomp (x) that satisfies the following equation.
ν

c j D j Scomp (x) = 0 . (3.58)
j=0
The sum of the complementary solution Scomp (x) and the particular integral I pi leads
to complete solution. That is, from (3.57) and (3.58), one has
ν

c j D j u(x) = B(x)
j=0
ν
ν

= c j D j Scs (x) = c j D i Scomp (x) + I pi
j=0 j=0
ν
ν

= c j D j Scomp (x) + c j D j I pi = 0 + B(x) = B(x) . (3.59)
j=0 j=0
3.3.1 Particular Integral: I pi
While the process of determining complementary solution for homogeneous linear

ordinary differential equation with constant coefficients has now been amply studied,
calculation of the particular integral, I pi , for inhomogeneous linear ordinary differ-
ential equations with constant coefficients still needs to be described. This matter is
dealt with in the current subsection. The procedure described here is a generalization
of what is known as the method of undetermined coefficients. Detailed explanation
is provided by working through several calculations that involve different versions
of the function B(x).
Calculation of I pi
ν

ci D i
I pi = B(x)
i=0
In order to solve and find the particular integral for an inhomogeneous linear ordinary
differential equation with constant coefficients given above, proceed as follows:
ν
i −1
Multiply both sides on the left by i=0 ci D . The result is
ν
−1 ν ν −1

ci D i
ci D i
I pi = I pi = ci D i
B(x) . (3.60)
i=0 i=0 i=0
ν
i −1
Expand i=0 ci D B(x) as series in ascending powers of D and retain only the
terms needed for the given B(x).
3.3.2 Examples Group V
I pi when B(x) = en x n + e0
V(A) : Solve

D 2 + D + 1 u(x) = e5 x 5 + e0 .
Calculate first the particular integral u(x) = I pi . To this end, as explained above,
invert the left-hand side and expand the resultant in ascending powers of D, retaining
terms only up to the order D 5 . Note that higher powers of D in the expansion would
contribute nothing.

1
V (A) : I pi = (e5 x 5 + e0 )
D2 + D + 1

= 1 − D + D 3 − D 4 + O(D 6 ) (e5 x 5 + e0 )
= e5 (x 5 − 5x 4 + 60x 2 − 120x) + e0 . (3.61)
Now calculate its complementary solution.

To that end, use the procedure previously described following (3.21) and (3.22).
That is, begin by trying exp (k x) as a solution.

D 2 + D + 1 exp(k x) = (k 2 + k + 1) exp(k x) = 0 . (3.62)
In the above equation, ignore exp(k x). The rest is the characteristic equation E ch .
(k 2 + k + 1) = 0 . (3.63)
The E ch here has complex roots: k = k1 = r + i m and k = k2 = r − i m, where

√
1 3
r =− ; m= . (3.64)
2 2
The resultant complementary solution consists of exp(k1 x) and exp(k2 x) and is
expressed in terms of
exp(r x) sin(mx) and exp(r x) cos(mx) . (3.65)
More precisely, it is
√ √
x 3 3
V (A) : Scomp (x) = exp − σ3 sin x + σ4 cos x .(3.66)
2 2 2
As usual, σ3 and σ4 are arbitrary constants. Also, the sign and the cosine functions
are linearly independent. As such (3.61) and (3.66) represent complete solution,
= I pi + Scomp (x), of differential equation V(A).
V(B): Solve

D 2 − D − 1 u(x) = (e3 x 3 + e0 ) .
The particular integral, as before, is calculated as follows:

1
V (B) : I pi = (e3 x 3 + e0 )
D2 − D − 1

= − 1 − D + 2D 2 − 3D 3 + O(D 4 ) (e3 x 3 + e0 )
= e3 (−x 3 + 3x 2 − 12x + 18) − e0 . (3.67)
Next proceed as in exercise V(A) and find the characteristic equation.
(k 2 − k − 1) = 0 . (3.68)
√
Its roots, k1,2 = 1
2
± 2
5
, lead to the complementary solution
√ √
x 5 5
V (B) : Scomp (x) = exp σ1 exp x + σ2 exp − x . (3.69)
2 2 2
V(C): Solve

D 2 + 2D + 1 u(x) = (e4 x 4 + e0 ) .
The particular integral is calculated the same way as in (3.61) and (3.67).

1
V (C) : I pi = (e4 x 4 + e0 )
D 2 + 2D + 1

= 1 − 2D + 3D 2 − 4D 3 + 5D 4 + O(D 5 ) (e4 x 4 + e0 )
= e4 (x 4 − 8x 3 + 36x 2 − 96x + 120) + e0 . (3.70)
Next, the Scomp (x). The characteristic equation k 2 + 2k + 1 = 0 has a double root
k = k1 = k2 = −1. Therefore, similar to (3.28)–(3.37), its complementary solution
is
V (C) : Scomp (x) = (σ0 + σ1 x) exp(−x) . (3.71)
A more detailed set of equations is solved in the following.
3.3.3 Examples Group VI
Solve the following twelve inhomogeneous linear ordinary differential equations

with constant coefficients. The homogeneous linear ordinary differential equations
with constant coefficients that occur in (3.55) have been converted here by adding
a term of the form B(x) = en x n + eo . Because Scomp (x) to these equations have
already been worked out—see (3.56)—only the particular integral, I pi , needs to
be worked out here. Complete solution can then be obtained by adding I pi to the
corresponding Scomp (x).

D 2 + 3D + 1 u(x) = (e1 x 3 + e0 ) . (1)

D 2 + 3D − 1 u(x) = (e2 x 3 + e0 ) . (2)

D 2 + 3D − 3 u(x) = (e3 x 3 + e0 ) . (3)

9
D + 3D +
2
u(x) = (e4 x 4 + e0 ) . (4)
2

D 2 + 4D + 6 u(x) = (e5 x 4 + e0 ) . (5)

1
D +D+
2
u(x) = (e6 x 4 + e0 ) . (6)
2

3 2
D+ u(x) = (e7 x 5 + e0 ) . (7)
2

3 2
D− u(x) = (e8 x 5 + e0 ) . (8)
2
(D + 2)2 u(x) = (e9 x 5 + e0 ) . (9)
(D + 1)3 u(x) = (e10 x 6 + e0 ) . (10)

D (D 2 − 4)2 u(x) = (e11 x 6 + e0 ) . (11)
2 2
D (D − 1) u(x) = (e12 x 6 + e0 ) . (12) (3.72)
Solution: (3.72)
−1
(1) I pi = D 2 + 3D + 1 (e1 x 3 + e0 )

= 1 − 3D + 8D 2 − 21D 3 + ... (e1 x 3 + e0 ) ,
= e1 (x 3 − 9x 2 + 48x − 126) + e0 .
−1
(2) I pi = D 2 + 3D − 1 (e2 x 3 + e0 )

= −1 − 3D − 10D 2 − 33D 3 + ... (e2 x 3 + e0 ) ,
= −e2 (x 3 + 9x 2 + 60x + 198) − e0 .

−1
(3) I pi = D 2 + 3D − 3 (e3 x 3 + e0 )

1 4 5
= −1 − D − D 2 − D 3 + ... (e3 x 3 + e0 ) ,
3 3 3
e e
3 3 2 0
=− x + 3x + 8x + 10 − .
3 3
−1
9
(4) I pi = D 2 + 3D + (e4 x 4 + e0 )
2

2 2 2 4 4
= 1 − D + D2 − D + ... (e4 x 4 + e0 ) ,
9 3 9 81

2 8 8 32 2
= e4 x4 − x3 + x2 − + e0 .
9 3 3 27 9
−1
(5) I pi = D 2 + 4D + 6 (e5 x 4 + e0 )

1 2 5 2 2 3 1 4
= 1− D+ D − D + D + ... (e5 x 4 + e0 ) ,
6 3 18 27 324

e5 4 8 3 10 2 16 2 e0
= x − x + x − x+ + .
6 3 3 9 27 6

1 −1
(6) I pi = D2 + D + (e6 x 4 + e0 )
2

= 2 e6 1 − 2D + 2D 2 − 4D 4 + ... x 4 + 2 e0 ,

= 2 e6 x 4 − 8x 3 + 24x 2 − 96 + 2 e0 .

3 −2
(7) I pi = D+ (e7 x 5 + e0 )
2

4 4 4 2 32 3 80 4 64 5
= 1− D+ D − D + D − D + ... (e7 x 5 + e0 ) ,
9 3 3 27 81 81

4 20 4 80 3 640 2 3200 2560 4
= e7 x 5 − x + x − x + x− + e0 .
9 3 3 9 27 27 9

3 −2
(8) I pi = D− (e8 x 5 + e0 )
2

4 4 4 32 3 80 4 64 5
= 1 + D + D2 + D + D + D + ... (e8 x 5 + e0 ) ,
9 3 3 27 81 81

4 5 20 4 80 3 640 2 3200 2560 4
= e8 x + x + x + x + x+ + e0 .
9 3 3 9 27 27 9
(9) I pi = (D + 2)−2 (e9 x 5 + e0 )

1 3 1 5 4 3 5
= 1 − D + D2 − D3 + D − D + ... (e9 x 5 + e0 ) ,
4 4 2 16 16

1 5 4 3 2 75 45 1
= e9 x − 5x + 15x − 30x + x− + e0 .
4 2 2 4
(10) I pi = (D + 1)−3 (e10 x 6 + e0 )

= 1 − 3D + 6D 2 − 10D 3 + 15D 4 − 21D 5 + 28D 6 − ... (e10 x 6 + e0 ) ,
= e10 (x 6 − 18x 5 + 180x 4 − 1200x 3 + 5400x 2 − 15120x + 20160) + e0 .

−1
(11) I pi = D (D 2 − 4)2 (e11 x 6 + e0 )

1
= + 8 D + 63 D 3 + 496 D 5 + O(D 7 ) (e11 x 6 + e0 )
D

x7 5 3
= e0 x + e11 + 48 x + 7560 x + 357120 x . (11)
7
−1
(12) I pi = D 2 (D 2 − 1) (e12 x 6 + e0 )

1
=− + 1 + D + D + D + O(D ) (e12 x 6 + e0 )
2 4 6 8
D2

x2 x8 6 4 2
= − e0 + 1 + e12 + x + 30x + 360x + 720 . (12) (3.73)
2 56
Note, in (11) above use was made of the relationship
1
f (x) = f (x) dx ,
D
and in (12)

1
f (x) = dx f (x) dx .
D2
Therefore
1 e11 7
(e11 x 6 + e0 ) = x + e0 x
D 7
and
1 e12 8 e0 2
2
(e12 x 6 + e0 ) = x + x .
D 56 2
In keeping with the tradition, I pi contains only known constants ei , etc. Unknown
constants such as σi , etc., are not included in the result above. That is appropriate
because such constants are already equivalently present in the relevant Scomp (x) in
(3.56) and therefore in the complete solution.
3.3.4 Problems Group II
Work out the particular integral I pi for the following ten inhomogeneous linear ordi-
nary differential equations with constant coefficients. The equations given below
were put together by making additions—in the form B(x) = cn x n + c0 —to the
homogeneous linear ordinary differential equation with constant coefficients given
in Problems Group I. Because the relevant complementary solutions are already
available from Problems Group I, only particular integrals, I pi , are being required
here.

D 2 + 2D − 3 u(x) = c1 x 3 + c0 . (1)

D 2 − 3D − 4 u(x) = c2 x 3 + c0 . (2)

1 2
D+ u(x) = c3 x 3 + c0 . (3)
2

1 2
D− u(x) = c4 x 4 + c0 . (4)
2
(D + 2)3 u(x) = c5 x 4 + c0 . (5)

(D − 2)3 u(x) = c6 x 4 + c0 . (6)

D 2 + D + 1 u(x) = c7 x 5 + c0 . (7)

D 2 − D + 1 u(x) = c8 x 5 + c0 . (8)

D 2 + 2D + 3 u(x) = c9 x 5 + c0 . (9)

D 2 + 3D + 4 u(x) = c10 x 6 + c0 . (10)
Procedure for Solving

I pi When B(x) = en exp(αx) + e0
In order to solve an inhomogeneous linear ordinary differential equation with con-

stant coefficients where B(x) = en exp(α x) + e0 , it is helpful to know the effect
of operating with a function of D on en exp(α x). To that purpose, consider the
following identity for integral n ≥ 0.
D n [en exp(α x)] = en αn exp(α x) . (3.74)
As such, when acting on exp(α x) a function f (D) that involves only positive powers
of D will lead to
f (D) [en exp(α x)] = en f (α) exp(α x) .
If f (α) = 0, one can divide both sides by f (α).
1 en f (α) exp(α x)
f (D) en exp(α x) = = en exp(α x). (3.75)
f (α) f (α)
From the left, multiply both sides of (3.75) by 1

f (D)
.
1 1 1
f (D) en exp(α x) = [en exp(α x)] . (3.76)
f (D) f (α) f (D)
And rewrite the left-hand side of (3.76) in an equivalent form.

1 1 en exp(α x)
f (D) en exp(α x) = . (3.77)
f (α) f (D) f (α)
The left-hand sides of (3.76) and (3.77) are equal and hence, also the right-hand
sides.
1 en exp(α x)
[en exp(α x)] = . (3.78)
f (D) f (α)
3.3.5 Examples Group VII
Here B(x) = en exp(α x) + e0

Solve the following twelve inhomogeneous linear ordinary differential equations with
constant coefficients. For convenience, the homogeneous part of these equations is the
same as (3.55). For calculating the relevant I pi , it is helpful to use (3.61), (3.67), (3.70)
and (3.78).

D 2 + 3D + 1 u(x) = e1 exp(α x) + e0 . (1)

D + 3D − 1 u(x) = e2 exp(α x) + e0 .
2
(2)

D + 3D − 3 u(x) = e3 exp(α x) + e0 .
2
(3)

9
D 2 + 3D + u(x) = e4 exp(α x) + e0 . (4)
2

D 2 + 4D + 6 u = e5 exp(α x) + e0 . (5)

1
D2 + D + u(x) = e6 exp(α x) + e0 . (6)
2

3 2
D+ u(x) = e7 exp(α x) + e0 . (7)
2

3 2
D− u(x) = e8 exp(α x) + e0 . (8)
2
(D + 2)2 u(x) = e9 exp(α x) + e0 . (9)
(D + 1)3 u(x) = e10 exp(α x) + e0 . (10)

D (D 2 − 4)2 u(x) = e11 exp(α x) + e0 . (11)
2 2
D (D − 1) u(x) = e12 exp(α x) + e0 . (12) (3.79)
Solution: I pi for (3.79)
1 e1 exp(α x) e0
I pi = (e1 exp(α x) + e0 ) = + . (1)
D 2 + 3D + 1 α2 + 3α + 1 1
1 e2 exp(α x) e0
I pi = (e2 exp(α x) + e0 ) = + . (2)
D 2 + 3D − 1 α2 + 3α − 1 −1
1 e3 exp(α x) e0
I pi = (e3 exp(α x) + e0 ) = + . (3)
D + 3D − 3
2 α + 3α − 3
2 −3
1 e4 exp(α x) e0
I pi = (e4 exp(α x) + e0 ) = + . (4)
D + 3D + 29
2 α + 3α + 2
2 9 9/2
1 e5 exp(α x) e0
I pi = (e5 exp(α x) + e0 ) = + . (5)
D 2 + 4D + 6 α2 + 4α + 6 6
1 e6 exp(α x) e0
I pi = (e6 exp(α x) + e0 ) = + . (6)
D2 + D + 21 α +α+ 2
2 1 1/2
1 e7 exp(α x) e0
I pi = 2 (e7 exp(α x) + e0 ) = + . (7)
3 2 9/4
D + 23 α+ 2
1 e8 exp(α x) e0
I pi = 2 (e8 exp(α x) + e0 ) = 2
+ . (8)
D − 23 α− 3 9/4
2
1 e9 exp(α x) e0
I pi = (e9 exp(α x) + e0 ) = + . (9)
(D + 2)2 (α + 2) 2 4
1 e10 exp(α x) e0
I pi = (e10 exp(α x) + e0 ) = + . (10)
(D + 1)3 (α + 1) 3 1
1 1
I pi = 2 (e11 exp(αx) + e0 ) = 2 [e11 exp(αx)]
D D −4 2 D D −42
−1 x
1 D2 e11 exp(α x)
+ 1− e0 = 2 + e0 . (11)
16 D 2 α α2 − 4 16
1 1
I pi = (e12 exp(αx) + e0 ) = [e12 exp(αx)]
D D2 − 1
2 D2D2 − 1
2
1 −1 e12 exp(α x) x
+ 1 − D2 e0 = 2 2 − e0 − e0 . (12) (3.80)
−D 2 α α −1 2
3.3.6 Problems Group III
The inhomogeneous linear ordinary differential equations with constant coefficients

given below are obtained by making additions—in the form B(x) = cn exp(αx) +
c0 —to the homogeneous linear ordinary differential equations with constant coeffi-
cients given in Problems Group I. Because the relevant complementary solutions are
already available from Problems Group I, only particular integrals, I pi , are needed
here.

D 2 + 2D − 3 u(x) = c1 exp(α x) + c0 . (1)

D 2 − 3D − 4 u(x) = c2 exp(α x) + c0 . (2)

1 2
D+ u(x) = c3 exp(α x) + c0 . (3)
2

1 2
D− u(x) = c4 exp(α x) + c0 . (4)
2
(D + 2)3 u(x) = c5 exp(α x) + c0 . (5)
(D − 2)3 u(x) = c6 exp(α x) + c0 . (6)

D 2 + D + 1 u(x) = c7 exp(α x) + c0 . (7)

D 2 − D + 1 u(x) = c8 exp(α x) + c0 . (8)

D 2 + 2D + 3 u(x) = c9 exp(α x) + c0 . (9)

D 2 + 3D + 4 u(x) = c10 exp(α x) + c0 . (10)
3.3.7 I pi for B(x) = cos(αx) ; sin(αx)
Procedure for Solving

In order to solve and find the particular integral for a given inhomogeneous linear
ordinary differential equation with B(x) = cos(αx), sin(αx), etc., use the following
relationships.
1
cos(αx) = exp(iαx) + exp(−iαx) ,
2
1
sin(αx) = exp(iαx) − exp(−iαx) . (3.81)
2i
This renders the inhomogeneous terms similar to those treated in detail in (3.79)
and (3.80).
3.3.8 Examples Group VIII
Use (3.81) and information provided in (3.78) to solve for the inhomogeneous linear
ordinary differential equations with constant coefficients given below.

D 2 + 3D + 1 u(x) = 2 cos(x) . (1)

D 2 + 3D − 1 u(x) = 2i sin(x) . (2)

D 2 + 3D − 3 u(x) = 2 cos(x) . (3)

9
D 2 + 3D + u(x) = 2 cos(x) . (4)
2

D 2 + 4D + 6 u(x) = 2i sin(x) . (5)

1
D2 + D + u(x) = 2i sin(x) . (6)
2

3 2
D+ u(x) = 4 cos2 (x) . (7)
2

3 2
D− u(x) = 4 sin2 (x) . (8)
2
(D + 2)2 u(x) = 2 cos(2x) . (9)

(D + 1)3 u(x) = 2i sin(2x) . (10)

D (D 2 − 4) u(x) = 4 cos2 (2x) . (11)
2 2
D (D − 1) u(x) = −4 sin2 (2x) . (12) (3.82)
I pi for (3.82)
Because the Scomp (x) have already been calculated for the differential equations
(3.82)—see (3.55) and (3.56)—only the I pi are worked out here.
2 1
(1) I pi : cos(x) = exp(i x) + exp(−i x)
D2+ 3D + 1 D + 3D + 1
2
exp(i x) exp(−i x) 2
− = sin(x) .
3i 3i 3
1 1
(2) I pi : 2i sin(x) = exp(i x) − exp(−i x)
+ 3D − 1
D2 D + 3D − 1
2

exp(i x) exp(−i x) −i
= − − = [4 sin(x) + 6 cos(x)] .
2 − 3i 2 + 3i 13
1 1
(3) I pi : 2 cos(x) = [exp(i x) + exp(−i x)]
+ 3D − 3
D2 D + 3D − 3
2

=− − = [−4 cos(x) + 3 sin(x)] .
4 − 3i 4 + 3i 25
1 1
(4) I pi : 2 cos(x) = [exp(i x) + exp(−i x)]
D2 + 3D + 2
9
D + 3D + 29
2

= 7 + 7 = [6 sin(x) + 7 cos(x)] .
2
+ 3i 2
− 3i 85
2i 1
(5) I pi : sin(x) = exp(i x) − exp(−i x)
+ 4D + 6
D2 D + 4D + 6
2

exp(i x) exp(−i x) i
= − = [10 sin(x) − 8 cos(x)] .
5 + 4i 5 − 4i 41
1 1
(6) I pi : [2i sin(x)] = exp(i x) − exp(−i x)
+D+2
D2 1
D +D+2
2 1

exp(i x) exp(−i x) 4i
= − = − [sin(x) + 2 cos(x)] .
3
2
+i 3
2
−i 5
1 1
(7) I pi : [4 cos (x)] =
2
2 [exp(2i x) + exp(−2i x) + 2]
3 2
D+2 D + 23
exp(2i x) exp(−2i x) 2 192 56 8
= + + 3 2 = sin(2x) − cos(2x) + .
3 2 3 2 625 625 9
2i + 2 −2i + 2 2
1 1
(8) I pi : [4 sin2 (x)] =
2 [2 − exp(2i x) − exp(−2i x)]
3 2
D− 2
D − 23
2 exp(2i x) exp(−2i x) 8 192 56
= 2 + + = + sin(2x) + cos(2x) .
3 2 3 2 9 625 625
3
2
2i + 2 −2i + 2

1 1 exp(2i x) exp(−2i x)
(9) I pi : [2 cos(2x)] = +
(D + 2)2 4 (1 + i)2 (1 − i)2
sin(2x)
= .
4
1 exp(2i x) − exp(−2i x)
(10) I pi : [2i sin(2x)] =
(D + 1) 3
(D + 1)3

exp(2i x) exp(−2i x) 2i
= − =− [11 sin(2x) − 2 cos(2x)] .
(1 + 2i)3 (1 − 2i)3 125
1 1
(11) I pi : [4 cos2 (2x)] = exp(4i x) + exp(−4i x)
D D2 − 4 D(D 2 − 4)

1 1 exp(4i x)
+ 2 =
D 3 − 4D 4i (4i)2 − 4
−1
1 exp(−4i x) 1 D2
− + 1− 2
4i (−4i)2 − 4 −4D 4
sin(4x) x
=− − .
40 2

−4 sin2 (2x) 1 exp(4i x) exp(−4i x)
(12) I pi : = 2 +
D2 D2 − 1 D D2 − 1 D2 − 1

1 1 exp(4i x)
− 2=
D4 − D2 (4i)2 (4i)2 − 1

1 exp(−4i x) 1 cos(4x)
+ + 2−2= + x2 − 2 . (3.83)
(4i)2 (−4i)2 − 1 D2 136
3.3.9 Problems Group IV
Additions have been made to the right-hand side in the form B(x) = sin(nx) and
cos(nx) to the homogeneous linear ordinary differential equations with constant
coefficients in Problems Group I. Work out the particular integrals, I pi , for the
following differential equations.

D 2 + 2D − 3 u(x) = 2 cos(x) . (1)

D 2 − 3D − 4 u(x) = 2 i sin(x) . (2)

1 2
D+ u(x) = sin(x) sin(2 x) . (3)
2

1 2
D− u(x) = sin(x) cos(2x) . (4)
2
(D + 2)3 u(x) = 4 cos2 (x) . (5)
(D − 2)3 u(x) = −4 sin2 (x) . (6)

D 2 + D + 1 u(x) = 2 cos(2x) . (7)

D 2 − D + 1 u(x) = 2i sin(2x) . (8)

D 2 + 2D + 3 u(x) = 4 cos2 (2x) . (9)

D 2 + 3D + 4 u(x) = −4 sin2 (2x) . (10)
3.3.10 I pi for B(x) = exp(α x) W (x)
Knowing
D m exp(α x) W (x) = exp(α x) (D + α)m W (x) ,
one can write
[(D)]n exp(α x) W (x) = {exp(α x)[(D + α)]n } W (x) (3.84)
where (D) is a function involving positive powers of D. Therefore, one calculates

the particular integral for B(x) = exp(α x) W (x) by following the same strategy as
employed in (3.75)–(3.78). That is, set n = −1.
I pi = [(D)]−1 exp(α x) W (x) = {exp(α x)[(D + α)]−1 } W (x)

exp(α x) exp(α x)
= W (x) = W (x) . (3.85)
[(D)] [(D + α)]
As a simple exercise, consider differential equation with B(x) = exp(x) 2x 3 .

D 2 + 2 D + 1 u = exp(x) 2x 3 . (3.86)
Notice α = 1 here.

1
I pi = 2 exp(x) 2x 3
D +2D+1

1
= exp(x) 2 x3
(D + 1)2 + 2 (D + 1) + 1
⎧ ⎫
exp(x) ⎨ 1 ⎬
= x 3
2 ⎩ 1 + D + D2 ⎭
4

exp(x) 3 2 1 3
= 1 − D + D − D + O(D ) x 3
4
2 4 2

exp(x) 9
I pi = x −3x + x −3 .
3 2
(3.87)
2 2
The characteristic equation, E ch , is found from the differential equation in the usual
manner. The complementary solution, Scomp (x), is found according to (3.39)–(3.42).
Both are given below in (3.88).
E ch : k2 + 2 k + 1 = 0 ;

D2 + 2 D + 1 u = 0 . (3.88)
The characteristic equation has double roots k = −1. Therefore, just as in (3.37),
the complementary solution is
Scomp (x) = exp(−x)(σ1 x + σ0 ) . (3.89)
The complete solution, Scs (x), is the sum of Scomp (x), (3.89), and the I pi , (3.87).
3.3.11 Examples Group IX
Again treat the same twelve homogeneous linear ordinary differential equations
with constant coefficients as in (3.55) and change them into inhomogeneous lin-
ear ordinary differential equations with constant coefficients by adding B(x) =
exp(α x) W (x). The above procedure can readily be generalized to deal with
cases where B(x) is either exp(α x){cos(x)}μ x n or exp(α x){sin(x)}μ x n . That
is so because B(x) can again be expressed in terms of such relationships as
exp(α x + ν) x n . In this spirit, solve the following set of twelve equations (3.90).

D 2 + 3D + 1 u(x) = exp(3x) x 2 . (1)

D 2 + 3D − 1 u(x) = exp(3x) x 2 . (2)

D 2 + 3D − 3 u(x) = exp(3x) x 2 . (3)

9
D + 3D +
2
u(x) = exp(4x) x 2 . (4)
2

D 2 + 4D + 6 u(x) = exp(4x) x 2 . (5)

1
D +D+
2
u(x) = exp(4x) x 2 . (6)
2

3 2
D+ u(x) = 2 exp(−x) cos(x) x . (7)
2

3 2
D− u(x) = 2 exp(x) cos(x) x . (8)
2
(D + 2)2 u(x) = 2i exp(−x) sin(x) x . (9)
(D + 1)3 u(x) = 2i exp(x) sin(x) x . (10)

D (D 2 − 4) u(x) = 4 cos2 (2x) x 2 . (11)
2 2
D (D − 1) u(x) = −4 sin2 (2x) x 2 . (12) (3.90)
I pi for (3.90)
We use (3.84), (3.85), and (3.86). Again, because Scomp (x) has already been calculated
for the differential equations (3.90)—see (3.55) and (3.56)—only the I pi are being
worked out here.
1 1
(1) I pi : exp(3x) x 2 = exp(3x) x2
+ 3D + 1
D2 (D + 3) + 3(D + 3) + 1
2

1 exp(3x) 9D 62D 2
= exp(3x) 2 x2 = 1− + x2
D + 9D + 19 19 19 361

exp(3x) 18 124
= x − x+
2
.(1)
19 19 361
1 1
(2) I pi : exp(3x) x 2 = exp(3x) x2
D 2 + 3D − 1 (D + 3)2 + 3(D + 3) − 1

1 exp(3x) 9D 64D 2
= exp(3x) 2 x 2
= 1− + x2
D + 9D + 17 17 17 289

exp(3x) 18 128
= x2 − x + .(2)
17 17 289
1 1
(3) I pi : exp(3x) x 2 = exp(3x) x2
D2 + 3D − 3 (D + 3) + 3(D + 3) − 3
2

1 exp(3x) 3D 22D 2
= exp(3x) 2 x2 = 1− + x2
D + 9D + 15 15 5 75

exp(3x) 6 44
= x − x+
2
.(3)
15 5 75
1 1
(4) I pi : exp(4x) x 2 = exp(4x) x2
D 2 + 3D + 9
(D + 4)2 + 3(D + 4) + 29
2

1 exp(4x) 22 354 2 2
= exp(4x) x =2
2
1− D+ D x
D2 + 11D + 65 65 65 (65)2
2

exp(4x) 44 708
=2 x2 − x + .(4)
65 65 (65)2
1 1
(5) I pi : exp(4x) x 2 = exp(4x) x2
+ 4D + 6
D2 (D + 4) + 4(D + 4) + 6
2

1 exp(4x) 6 53 2 2
= exp(4x) 2 x2 = 1− D+ D x
D + 12D + 38 38 19 722

exp(4x) 12 53
= x − x+
2
.(5)
38 19 (19)2
1 1
(6) I pi : exp(4x) x 2 = exp(4x) x2
D 2 + D + 21 (D + 4)2 + (D + 4) + 21

1 exp(4x) 18 242 2 2
= exp(4x) 2 x =2
2
1− D+ D x
D + 9D + 41 41 41 (41)2
2

exp(4x) 36 484
=2 x2 − x + .(6)
41 41 (41)2
1
(7) I pi : 2 [exp(−x) 2 cos(x) x]
D + 23
1 1
= exp{x(−1 − i)} 2 x + exp{x(−1 + i)} 2 x
D − 1 − i + 23 D − 1 + i + 23

exp{−x(1 + i)} 4 exp{−x(1 − i)} 4
= 1 + (3 − 4i)D x + 1 + (3 + 4i)D x
− 43 − i 25 − 43 + i 25

8x exp(−x) 32 exp(−x)
= [4 sin(x) − 3 cos(x)] + [2 sin(x) + 11 cos(x)] . (7)
25 125
1
(8) I pi : [exp(x) 2 cos(x) x]
3 2
D− 2
1 1
= exp{x(1 + i)} x + exp{x(1 − i)} 2 x
3 2
D+1+i − D + 1 − i − 23
2

exp{x(1 + i)} 4 exp{x(1 − i)} 4
= 1 + (1 + 2i)D x + 1 + (1 − 2i)D x
− 43 − i 5 − 43 + i 5

8x exp(x) 8 exp(x)
=− [4 sin(x) + 3 cos(x)] + [8 sin(x) − 44 cos(x)] . (8)
25 125
1
(9) I pi : [exp(−x) 2i sin(x) x]
(D + 2)2
1 1
= exp{−x(1 − i)} x − exp{−x(1 + i)} x
(D − 1 + i + 2) 2
(D − 1 − i + 2)2
exp{−x(1 − i)} exp{−x(1 + i)}
= [1 + (i − 1)D]x + [1 − (i + 1)D]x
2i 2i
= i exp(−x)[sin(x) + cos(x) − x cos(x)] . (9)
1
(10) I pi : [exp(x) 2i sin(x) x]
(D + 1)3
1 1
= exp{x(1 + i)} x − exp{x(1 − i)} x
(D + 1 + i + 1) 3
(D + 1 − i + 1)3

exp{x(1 + i)} 3D exp{x(1 − i)} 3D
= 1− x− 1− x
(2 + i) 3 2 + i (2 − i) 3 2 −i
2i exp(x)
= [21 sin(x) + 72 cos(x) + 10 x sin(x) − 55 x cos(x)] . (10)
625
1 1
(11) I pi : [4 cos2 (2x) x 2 ] = exp(4i x) + exp(−4i x) + 2 x 2
D D2 − 4 D 3 − 4D
1 1
= exp(4i x) x 2 + exp(−4i x) x2
(D + 4i)3 − 4(D + 4i) (D − 4i)3 − 4(D − 4i)
⎡ ⎤
1 1
+⎣ 2
⎦ 2x 2 = exp(4i x)
3 + 12i D 2 − 52D − 80i
x2
−4D 1 − 4 D D

1 − D 2 −1
exp(−4i x) 4
+ 3 x2 + 2x 2
D − 12i D 2 − 52D + 80i −4D

i 13 109i 2 2
= exp(4i x) − D− D x +
80 1600 32000

−i 13 109i 2 2 x x3
exp(−4i x) − D+ D x − −
80 1600 32000 4 6
2 sin(4x) cos(4x)
= −x − 13 x
40 400

109 x x3
+ sin(4x) − − . (11)
8000 4 6
1 1
(12) I pi : [−4 sin2 (2x)x 2 ] = exp(4i x) + exp(−4i x) − 2 x 2
D2 D2 − 1 D4 − D2
1 1
= exp(4i x) x 2 + exp(−4i x) x2
(D + 4i) − (D + 4i)
4 2 (D − 4i) − (D − 4i)2
4

1
+ 2x 2 =
D2 1 − D2
1
exp(4i x) × x2
D 4 + 16i D 3 − 97D 2 − 264i D + 272
1
+ exp(−4i x) × 4 x2
D − 16i D 3 − 97D 2 + 264i D + 272
1 −1
+ 2 1 − D2 2x 2 =
D
exp(4i x) 33 2707 2 2
1+ i D− D x
272 34 4624

exp(−4i x) 33 2707 2 2 x 4
+ 1− i D− D x + + 2x 2 + 4 =
272 34 4624 6

cos(4x) 2707 33
x2 − − x sin(4x)
136 2312 17 × 136
x4
+ + 2x 2 + 4. (12) (3.91)
6
3.3.12 Problems Group V
On the right-hand side of the homogeneous linear ordinary differential equations

with constant coefficients—see Problems Group I—additions have been made in the
form B(x) = exp(α x) W (x) where W (x) ≡ sin(x)x n , or cos(x) x n , etc. Work out
their particular integral, I pi .

D 2 + 2D − 3 u(x) = 2 exp(2x) cos(x) x . (1)

D 2 − 3D − 4 u(x) = 2 i exp(2x) sin(x) x . (2)

1 2
D+ u(x) = sin(x) cos(2x) . (3)
2

1 2
D− u(x) = sin(x) sin(2x) . (4)
2
(D + 2)3 u(x) = 4 exp(x) cos(x) x . (5)
(D − 2)3 u(x) = 4 exp(x) sin(x) x . (6)

D + D + 1 u(x) = −4 exp(x) sin2 (x) x . (7)
2

D 2 − D + 1 u(x) = 4 exp(x) cos2 (x) x . (8)

D 2 + 2D + 3 u(x) = 4 x cos(x) exp(−x) . (9)

D 2 + 3D + 4 u(x) = −4 x sin(x) exp(x) . (10)
In particular, for problems (3) and (4), it is helpful to note the equalities sin(x)
cos(2 x) = sin(x) − 2 sin3 (x) and sin(x) sin(2 x) = 2 cos(x) − 2 cos3 (x).
3.4 Simultaneous Linear (ODEs) with Constant

Coefficients
An (ODE) has only one independent variable and usually only a single dependent
variable. But occasionally, the needs of a subject-matter require more than one depen-
dent variable. For complete description then, there must be either partial differential
equations or more than one simultaneous linear ordinary differential equation with
constant coefficients. Here, we treat the latter option. Generally, the larger the num-
ber of these equations, the greater the effort needed to solve them. Therefore, for
convenience, we work with equations that have only one, two, or three dependent
variables. Because there are cases where the differential equations can be separated
so that each is a function only of one dependent variable. Therefore, for additional
convenience, we work with such cases first.
3.4.1 Separable Cases
Chose t as the single independent variable and notate a first differential with respect
to t by the symbol . That is
d
≡ .
dt
{A} Number of Constants
{A}: Consider first a very simple example with only two dependent variables y ≡
y(t) and z ≡ z(t) that satisfy the following simultaneous equations.
y(t) − z(t) = 0 ,
y(t) − z(t) = 0 . (3.92)
As noted in the passage following (3.4), the minimum number of constants needed for
a complete solution to a single homogeneous linear ordinary differential equation
with constant coefficients is equal to the order of the given differential equation.
But when dealing with a series of coupled homogeneous linear ordinary differential
equations with constant coefficients, the answer to this question is more subtle. Here,
one needs to work with the determinant of the coefficients that multiply each of the
3.4 Simultaneous Linear (ODEs) with Constant Coefficients 47
dependent variables and then look for the highest power of the differentials that occur.
A practical demonstration helps with explanation of this statement.
The determinant of the operational coefficients of the dependent variables y(t)
and z(t) in the two coupled equations (3.92) is

−1
= − 2 + 1 .
1 −
Because this determinant is not manifestly equal to zero, look for the highest power
of that occurs. It is equal to two. Therefore, the complete solution to this pair of
differential equations will have two arbitrary constants.
{A} Solution
The dependent variables y(t) and z(t) in (3.92) can readily be separated. To this end,
operate by from the left on (3.92).
[y(t) − z(t)] = 0 ; [y(t) − z(t)] = 0 . (3.93)
Now, by using the original equation, (3.92), eliminate −z(t) and y(t).
[y(t)] − y(t) = 0 ; [z(t)] − z(t) = 0 . (3.94)
The result is two homogeneous linear ordinary differential equations with constant
coefficients. Each of these two differential equations involves only a single dependent
variable, that is, y(t) or z(t). And they both have constant coefficients. Now the well-
worked procedure can be used to find their solution.
E ch; y : k 2 − 1 = 0 : k1,2 = ± 1 ; E ch; z : k 2 − 1 = 0 : k1,2 = ± 1.

Scomp; y (t) = σ1 exp(−t) + σ2 exp(t) ; Scomp; z (t)
= σ3 exp(−t) + σ4 exp(t). (3.95)
Unfortunately, the full complementary solution, Scomp; y (t) + Scomp; z (t), contains
four arbitrary independent constants: σ1 → σ4 . But there should really be a total
of only two undetermined constants. So what to do? Substitute the results, namely
y(t) = Scomp; y (t) and z(t) = Scomp; z (t), into original differential equation (3.92) and
see what happens. One gets
Scomp; y (t) = Scomp; z (t) ; Scomp; z (t) = Scomp; y (t) . (3.96)
Good. This yields the required equalities
− σ1 exp(−t) + σ2 exp(t) = σ3 exp(−t) + σ4 exp(t) ;

−σ3 exp(−t) + σ4 exp(t) = σ1 exp(−t) + σ2 exp(t) . (3.97)
And satisfaction of these equalities requires

σ3 = − σ1 ; σ4 = σ2 . (3.98)
Therefore, according to (3.95) and (3.98), the solution to the simultaneous linear
ordinary differential equations with constant coefficients (3.92) is the following:
Scomp; y (t) = σ1 exp(−t) + σ2 exp(t) ;

Scomp; z (t) = − σ1 exp(−t) + σ2 exp(t) . (3.99)
{B}: The simultaneous linear ordinary differential equations with constant coef-
ficients {A} in (3.92) were easy to solve. Let us continue to treat similarly simple
simultaneous equations but increase their number to three.
x(t) = y(t) ;
y(t) = z(t) ;
z(t) = x(t) . (3.100)
{B}: Number of Independent Constants

It is convenient to rewrite (3.100).
x(t) − y(t) + 0 × z(t) = 0 ,

0 × x(t) + y(t) − z(t) = 0 ,
x(t) + 0 × y(t) − z(t) = 0 .
There are three dependent variables x(t), y(t), and z(t) with t as the independent
variable. The determinant of the operational coefficients of the given simultaneous
equations is

−1 0

0 − 1 .

1 0
Again because this determinant is not manifestly equal to zero, the relevant parameter
is the highest power of that occurs. It is equal to three. Therefore, the complete
solution will have three arbitrary constants.
{B}: Solution
Once again the dependent variables can readily be separated. To separate x(t), from
y(t) and z(t), operate by from the left on the top equation in (3.100).
[x(t)] = y(t) . (3.101)
Next operate by on the middle equation in (3.100) and make use of the bottom
equation there.
[y(t)] = z(t) = x(t) . (3.102)
Combining (3.101) and (3.102) leads to a differential equation in single dependent

variable x(t). That is
(3 − 1)x(t) = 0 . (3.103)
In fact, the same process can be repeated for the dependent variables y(t) and z(t).
As a result, two remaining equations are obtained. Because of the x(t), y(t), z(t)
symmetry inherent in (3.100), the last two equations are very much like the first
differential equation (3.103). That is
(3 − 1)y(t) = 0 ;
(3 − 1)z(t) = 0 . (3.104)
All these three homogeneous linear ordinary differential equations with constant
coefficients are of third order. Therefore, complete solution of each will have three
independent constants, thus making a grand total of nine constants. But as noted
before, only three are needed. Therefore, six unnecessary constants will have to be
eliminated. Fortunately, because the given simultaneous linear ordinary differential
equations are all symmetric in structure, one needs to solve only one of the three.
The Scomp; x (t) for (3.103) is found in the usual manner as follows.
√
−1 ± i 3
E ch; x : (k) − 1 = 0 : k1 = 1 ; k2,3
3
= .
2
Therefore,
√ √
t 3 3
Scomp; x (t) = σ1 exp(t) + exp − σ2 exp −i t + σ3 exp i t
2 2 2
√ √
t 3 3
= σ1 exp(t) + exp − (σ2 + σ3 ) cos t − i(σ2 − σ3 ) sin t . (3.105)
2 2 2
Because of symmetry, the other two Scomp s can be written by inspection.

√ √
t 3 3
Scomp; y (t) = σ4 exp(t) + exp − σ5 exp −i t + σ6 exp i t
2 2 2
√ √
t 3 3
= σ4 exp(t) + exp − (σ5 + σ6 ) cos t − i(σ5 − σ6 ) sin t ;
2 2 2
√ √
t 3 3
Scomp; z (t) = σ7 exp(t) + exp − σ8 exp −i t + σ9 exp i t
2 2 2
√ √
t 3 3
= σ7 exp(t) + exp − (σ8 + σ9 ) cos t − i(σ8 − σ9 ) sin t . (3.106)
2 2 2
To eliminate the six unnecessary constants, substitute the results provided in (3.105)
and (3.106) in the form: x(t) = Scomp; x (t), y(t) = Scomp; y (t), and z(t) = Scomp; z (t),
into the original differential equations (3.100). One gets
Scomp; x (t) = Scomp; y (t) ; (3.107)

Scomp; y (t) = Scomp; z (t) ; (3.108)
Scomp; z (t) = Scomp; x (t) . (3.109)
These equations provide the relationships that allow for the cancelation of six of
the nine unknowns. But despite the straightforward nature of the needed algebra,
direct elimination of these constants from (3.107) to (3.109), as they are currently
structured, is a lengthy undertaking. With a view to finding a more convenient format
for these equations that would help relieve this difficulty, rewrite Scomp; x (t) in (3.105)
as follows.

t
Scomp; x (t) = σ1 exp(t) + σ0 exp −
2
√ √
3 3
× cos(φ) cos t − sin(φ) sin t .
2 2
Or equivalently
√
t 3
Scomp; x (t) = σ1 exp(t) + σ0 exp − cos t +φ . (3.110)
2 2
Here,

σ2 + σ3 σ2 − σ3
cos(φ) = ; sin(φ) = i . (3.111)
σ0 σ0
Because cos2 (φ) + sin2 (φ) = 1, the constant σ0 is chosen such that

σ2 σ3
4 =1 . (3.112)
σ02
Note that Scomp; x (t) still has three arbitrary constants. In addition to σ1 , there are the
two new constants σ0 and the angle φ. Of course, σ0 and φ are functions of the original
arbitrary constants σ2 and σ3 . Again, because of symmetry already mentioned, the
structure of Scomp; y (t) and Scomp; z (t) can be predicted by inspection.
√
t 3
Scomp; y (t) = σ1 exp(t) + σ0 exp − cos t +φ
: (a)
2 2
√
t 3
Scomp; z (t) = σ1 exp(t) + σ0 exp − cos t +φ
: (b). (3.113)
2 2
At this juncture, get back to the original equations and work with them in the
form (3.107)–(3.109). To this end, first use Scomp; x (t) given in (3.110) and differen-
tiate it to calculate Scomp; x (t).
Scomp; x (t) = σ1 exp(t)

√ √ √
t 1 3 3 3
+ σ0 exp − − cos t +φ − sin t +φ ,
2 2 2 2 2
√
t 3 2π
= σ1 exp(t) + σ0 exp − cos t +φ+ .
2 2 3
√3
(Trigonometric equalities, cos 2π3
= − 21 and sin 2π3
= 2 , were used here.)
Next write the result according to (3.107) as Scomp; y (t) = Scomp; x (t). That is
√
t 3
σ1
exp(t) + σ0
exp − cos t +φ
2 2
√
t 3 2π
= σ1 exp(t) + σ0 exp − cos t +φ+ .
2 2 3
This equation suggests the equalities:
σ1 = σ1 ,
σ0 = σ0 ,
2π
φ = φ + .
3
Accordingly,
√

t 3 2π
Scomp; y (t) = σ1 exp(t) + σ0 exp − cos t +φ+ . (3.114)
2 2 3
Finally, because of symmetry, (3.110) and (3.114) lead by induction to the final
result
√
t 3 4π
Scomp; z (t) = σ1 exp(t) + σ0 exp − cos t +φ+ . (3.115)
2 2 3
{C}: The simultaneous differential equations in problem {B} were in principle ele-
mentary even though eliminating the six unnecessary independent constants took
effort. In (3.116) given below, the level of complexity is raised a little. But here one
is helped by the fact that there are fewer constants that need eliminating.
( + 1)x(t) − 2y(t) = exp(t) ; − x(t) + 2( + 1)y(t) = exp(t).(3.116)
{C}: Number of Independent Constants

The independent variable is still t but now there are only two dependent variables
x(t) and y(t). The determinant of the operational coefficients is

( + 1) −2
.
−1 2( + 1)
of that occurs. Clearly, it is equal to two. Therefore, the complete solution to
this pair of simultaneous linear ordinary differential equations will have only two
arbitrary constants.
{C}: Solution
Following the usual protocol, it is necessary to remove one dependent variable from
each of the given two simultaneous linear ordinary differential equations. Only ele-
mentary algebra is needed first to eliminate y and next x. The result is the following
pair of differential equations.
(2 + 2)x(t) = 3 exp(t) ;

2(2 + 2)y(t) = 3 exp(t) . (3.117)
Using the established routine, (3.117) are straightforward to solve.
E ch; x : k 2 + 2k = 0 ; k1 = 0 ; k2 = −2 ;
E ch; y : 2(k 2 + 2k) = 0 ; k1 = 0 ; k2 = −2 ;
Scomp; x (t) = σ1 + σ2 exp(−2 t) ; Scomp; y (t) = σ3 + σ4 exp(−2 t) ;
1 3
I pi ; x = exp(t) .3 = exp(t) ;
( + 1)2 + 2( + 1) 3
1 3
I pi ; y = exp(t) .3 = exp(t) .(3.118)
2[( + 1) + 2( + 1)]
2 6
Once again there are more independent constants—meaning σ1 → σ4 —than are

needed. To eliminate the unnecessary additional constants, substitute the results for
x = Scomp; x (t) + I pi ; x and y = Scomp; y (t) + I pi ; y into at least one of the original
equations. The first of the original equation (3.116), namely ( + 1) x(t) − 2 y(t)
= exp(t), gives
( + 1) I pi ; x + Scomp; x (t) − 2 I pi ; y + Scomp; y (t) = exp (t) . (3.119)
That is
( + 1) {exp(t) + σ1 + σ2 exp(−2 t)}

1
−2 exp(t) + σ3 + σ4 exp(−2 t) = exp(t) . (3.120)
2
Or equivalently
(σ1 − 2σ3 ) − (σ2 + 2σ4 ) exp(−2 t) = 0 . (3.121)
For arbitrary t, this equation can be satisfied only if σ3 = σ21 and σ4 = − σ22 . Hence,
the complete solution of the simultaneous differential equations (3.116) is
x(t) = σ1 + σ2 exp(−2 t) + exp(t) ;

σ1 σ2 1
y(t) = − exp(−2 t) + exp(t) . (3.122)
2 2 2
{D}: Simultaneous linear differential equations (3.123) given below will be solved
next.
( + 1)2 x(t) + ( + 2)2 y(t) = exp(t) ;

( + 1)x(t) + ( + 2)y(t) = exp(t) t . (3.123)
{D}: Number of Independent Constants

Again, there are two dependent variables x and y and t is the independent variable.
The determinant of the operational coefficients is

( + 1)2 ( + 2)2

( + 1) ( + 2) .
of that occurs. It is equal to two. Therefore, the complete solution to this pair of
differential equations will have two arbitrary constants.
{D}: Solution
To eliminate x(t), multiply the bottom line in (3.123) by ( + 1) and subtract the
result from the line above.
( + 2)2 y(t) − ( + 1)( + 2)y(t) = exp(t) − ( + 1) exp(t) t .(3.124)
Equivalently,
( + 2)y(t) = −2 t exp(t) . (3.125)
Following the usual routine, (3.125) gives:
E ch; y : k + 2 = 0 ; k1 = −2 ; Scomp; y = σ1 exp(−2 t) ;

1 1
I pi ; y = exp(t)(−2 t) = exp(t) (−2 t)
() + 2 ( + 1) + 2

1 exp(t) 2
= exp(t) 1− (−2 t) = −2 t + . (3.126)
3 3 3 3
Clearly, σ1 is one of the two independent constants. This leaves only one additional
independent constant to find.
Next, in order to eliminate y(t), multiply the bottom line in (3.123) by ( + 2)
and subtract the result from the top line.
( + 1)2 x9t) − ( + 2)( + 1)x = exp(t) − ( + 2) [exp(t) t] .
Equivalently,
( + 1)x(t) = 3 t exp(t) . (3.127)
Again following the usual routine, one gets
E ch; x : k + 1 = 0 ; k1 = −1 ; Scomp; x (t) = σ2 exp(− t) ;

1 1
I pi ; x = exp(t)(3 t) = exp(t) (3 t)
() + 1 ( + 1) + 1

1 exp(t) 3
= exp(t) 1− (3 t) = 3t − . (3.128)
2 2 2 2
{E}: Solve
(2 + 3 + 4)x(t) + (2 + + 4)y(t) = exp(t) t ,

(2 + 1)x(t) + ( + 1)y(t) = 2 t . (3.129)
{E}: Number of Independent Constants

Again, t is the independent variable and there are two dependent variables x(t) and
y(t). The determinant of their operational coefficients is
2
( + 3 + 4) (2 + + 4)
.
(2 + 1) ( + 1)
Because this determinant is not manifestly equal to zero, look for the highest power of
that occurs. It is equal to three. Therefore, the complete solution to the simultaneous
linear ordinary differential equation (3.129) will have three arbitrary constants.
{E}: Solution
To eliminate x(t), multiply the top line in (3.129) by (2 + 1) and the bottom by
(2 + 3 + 4). Subtract the resultant equation at the bottom from the corresponding
one at the top.
(2 − + 2)y(t) = (2 + 1) exp(t) t − (2 + 3 + 4){2 t} .
Equivalently,
(2 − + 2)y(t) = 3 exp(t) t + 2 exp(t) − 8 t − 6 . (3.130)
Following the usual routine, (3.130) gives:

√
1±i 7
E ch; y : k(k − k + 2) = 0 ; k1 = 0 , k2,3 =
2
;
√ √ 2
t 7 7
Scomp; y (t) = σ1 + exp σ2 sin t + σ3 cos t ;
2 2 2
1
I pi ; y = {exp(t)(3 t + 2) − 8 t − 6}
(2 − + 2)
1
= exp(t) {3 t + 2}
( + 1)[( + 1)2 − ( + 1) + 2]
1
− (8 t + 6)
(2 − + 2)

1 3 1 1 1
= exp(t) 1 − {3 t + 2} − + − (8 t + 6) .
2 2 2 2 2

exp(t) 5 1
I pi ; y = 3t − − 2 t2 − 5 t + . (3.131)
2 2 2
Similarly, to eliminate y(t) multiply the top line in (3.129) by ( + 1) and the bottom
by (2 + + 4). Subtract the resultant equation at the top from the corresponding
one at the bottom.
(2 − + 2)x(t) = (2 + + 4){2 t} − ( + 1) exp(t) t ,

= −2 exp(t) t − exp(t) + 8 t + 2 . (3.132)
Equivalently,
(2 − + 2)x = −2 exp(t) t − exp(t) + 8 t + 2. (3.133)

According to the well-established routine, (3.133) gives

√
1±i 7
E ch; x : k(k − k + 2) = 0 ; k1 = 0 , k2,3 =
2
;
2
√ √
t 7 7
Scomp; x (t) = σ4 + exp σ5 sin t + σ6 cos t ;
2 2 2
1
I pi ; x = {exp(t)(−2 t − 1) + 8 t + 2}
(2 − + 2)

1
= exp(t) {−2 t − 1}
( + 1)[( + 1)2 − ( + 1) + 2]

1
+ (8 t + 2)
(2 − + 2)

1 3 1 1
= exp(t) 1 − {−2 t − 1} + + − (8 t + 2) .
2 2 2 4 4
3
I pi ; x = exp(t) (1 − t) + 2 t 2 + 3 t − . (3.134)
2
As usual there are more independent constants, meaning six—that is, σ1 → σ6 —
instead of the only three that are needed. And again, to eliminate the unnecessary
additional constants, substitute the results for x(t) = Scomp; x (t) + I pi ; x and y(t) =
Scomp; y (t) + I pi ; y —that is
3
x(t) = exp(t) (1 − t) + 2 t 2 + 3 t −
2
√ √
t 7 7
+ σ4 + exp σ5 sin t + σ6 cos t ;
2 2 2

exp(t) 5 1
y(t) = 3t − − 2 t2 − 5 t +
2 2 2
√ √
t 7 7
+ σ1 + exp σ2 sin t + σ3 cos t , (3.135)
2 2 2
into at least one of the original equations. The second of the original equa-
tion (3.123)—namely (2 + 1) x(t) + ( + 1)y(t) − 2 t = 0—is clearly the eas-
ier to handle. After a little algebra, one gets
√ √
t 7 3 7
(σ1 + σ4 ) + exp sin t (σ2 + σ5 ) − (σ3 + σ6 )
2 2 2 2
√ √
t 7 3 7
+ exp cos t (σ3 + σ6 ) + (σ2 + σ5 ) = 0 . (3.136)
2 2 2 2
For arbitrary t, this equation can be satisfied only if
(σ1 + σ4 ) = 0 ; (σ2 + σ5 ) = 0 ; (σ3 + σ6 ) = 0 . (3.137)
Insert σ4 = −σ1 , σ5 = −σ2 and σ6 = −σ3 into (3.135) describing x(t). One gets
3
x(t) = exp(t) (1 − t) + 2 t 2 + 3 t −
2
√ √
t 7 7
− σ1 − exp σ2 sin t + σ3 cos t . (3.138)
2 2 2
Equations (3.135) for y(t) and (3.138) for x(t) are the complete solution to the
pair of simultaneous linear ordinary differential equation (3.129) and appropriately
have only three arbitrary constants σ1 , σ2 , and σ3 .
{F}: Solve
()x(t) + (4 )y(t) = σ0 t ,
(3 )x(t) + (12 )y(t) = 2 t , (3.139)
where t is the independent variable and x(t) and y(t) are dependent variables.
{F}: Number of Independent Constants
The determinant of the operational coefficients of (3.139) is

() (4)

(3) (12) = 0. (3.140)
In other words, the relevant determinant is manifestly equal to zero. Thus, one won-
ders as to whether a solution to the simultaneous linear ordinary differential equa-
tion (3.139) is at all possible? And, if a solution is possible, how many independent
constants would there be?
In an attempt to deal with these issues, let us separate the equations. To this end,
multiply the first of the (3.139) from the left with (12) and the second with (4)
and subtract the second from the first, thereby canceling the terms multiplying y(t).
One gets
[(12) × () − (4) × (3)]x(t) = 0 = (12)(σ0 t) − (4)(2 t)

= 12 σ0 − 8 (3.141)
Therefore, in order that there be a solution to (3.139), the following has to be true
2
σ0 = . (3.142)
3
And if this relationship should be true, then the two equations in (3.139) are identical

and one is left with only a single equation: namely ()x(t) + (4 )y(t) = 23 t.
And by itself this equation contains insufficient information to solve for both x(t)
and y(t). On the other hand, if one of the independent variables—say y(t)—was
chosen, the differential equation involving x(t) would very likely have a solution.
But because the choice of y(t) is arbitrary, it should in principle have an arbitrary
number of constants. Thus, there is no fixed limit as to how many independent
constants such a solution would have.
3.4.2 Problems Group VI
Solve the following pairs of simultaneous linear ordinary differential equations with
constant coefficients. Reminder: = dtd and D = dx d
.
( + 1)x(t) + ( + 2)y(t) = exp(t) ,

(2 + 1)x(t) + (4 + 3)y(t) = exp(t) t . (1)
(2 + 1)x(t) + ( + 2)y(t) = exp(t) t ,
( + 2)x(t) + (2 + 1)y(t) = t . (2)
(2 + + 1)x(t) + ( + 1)y(t) = t ,
(2 + )x(t) + ( + 2)y(t) = t exp(t) . (3)
1
[(−x + 3)D + 1] u(x) + [(−x + 3)D + 2] v(x) = − ,
x −3
1
[2(−x + 3)D + 1] u(x) + [4(−x + 3)D + 3] v(x) = log (−x + 3) . (4)
x −3
1 '
[2(2x − 1)D + 1] u(x) + [(2x − 1)D + 2] v(x) = log (2x − 1) (2x − 1) ,
2
1
[(2x − 1)D + 2] u(x) + [2(2x − 1)D + 1] v(x) = log (2x − 1) . (5)
2

(x + 1)2 D 2 + 2(x + 1)D + 1 u(x) + [(x + 1)D + 1] v(x) = log (x + 1) ,

(x + 1)2 D 2 + 2(x + 1)D u(x) + [(x + 1)D + 2] v(x) = (x + 1) log (x + 1) . (6) (3.143)
Chapter 4
Variable Coefficients
Both single and simultaneous linear ordinary differential equations with constant
coefficients were discussed in detail in Chap. 3. Here, that analysis is extended to lin-
ear ordinary differential equations with variable coefficients. Differential equations
with non-constant coefficients are in general much harder to solve. For simplicity,
therefore, only first-order and first-degree equations of type (A) are handled.
(A) : Du(x) + M(x)u(x) = N (x) .
These equations are treated by using an appropriate integrating factor.

Included also is an extensive discussion of equations which outwardly look fear-
some: that is, equations of type
Du(x) + M(x)u(x) = N (x)u n (x) .
But with help from the Ber nouilli 1. suggestion, these equations can be transformed
into equations similar to (A).
4.1 Linear (ODE)’s
4.1.1 First-Order and First-Degree
Linear ordinary differential equations that are of the first-order and first-degree and
have variable coefficients are of the general form
f 1 (x)Du(x) + f 2 (x)u(x) = f 3 (x) . (4.1)

https://doi.org/10.1007/978-3-319-76406-1_4
60 4 Variable Coefficients
More conveniently, they are written as
Du(x) + M(x)u(x) = N (x) (4.2)

f 3 (x) f 2 (x)
where N (x) = f 1 (x)
and N (x) = f 1 (x)
and D = d
dx
.
4.1.2 Integrating Factor
Multiply (on the left) the differential equation (4.2) by an as yet unknown factor
F(x).
F(x)[Du(x) + M(x) u(x)] = F(x)N (x) . (4.3)
Assume F(x) is so chosen that it admits of the following equality
F(x)[Du(x) + M(x) u(x)] = D[F(x) u(x)] . (4.4)
This implies
F(x) [Du(x)] + [F(x) M(x)] u(x) = F(x) [Du(x)] + u(x) [D F(x)] . (4.5)
Remove F(x)[Du(x)] from both sides of the relationship (4.5).
[F(x) M(x)] u(x) = u(x) [D F(x)] . (4.6)
Notice the function u(x) is not operated upon either side of (4.6). Indeed, it ap-
pears simply as a multiplying factor. Therefore, it can be eliminated. The result is a
differential equation obeyed by the as yet unknown factor F(x).
F(x) M(x) = D F(x) . (4.7)

1
Multiply (from the left) by F(x)
and integrate with respect to x.

1
F(x)M(x) . dx = M(x) . dx
F(x)

1 1
≡ . D F(x) . dx = . dF(x) = log F(x) − log σ0 .
F(x) F(x)
Or equivalently

log F(x) = log σ0 + M(x) . dx . (4.8)
4.1 Linear (ODE)’s 61
Exponentiation determines the factor F(x).
F(x) = σ0 W (x) , (4.9)
where,
W (x) = exp M(x) dx (4.10)
and σ0 is an arbitrary constant. Henceforth, W (x) will be referred to as the integrating

factor. Clearly, W (x) is a known function because it depends only on M(x) which is
available as part of the relevant differential equation itself. [For instance, see (4.2).]
4.1.3 Equation (4.2): Solution
Left-hand sides of (4.3) and (4.4) are the same. Therefore, their right-hand sides must
be equal.
F(x)N (x) = D[F(x) u(x)] . (4.11)
In (4.11), as per (4.9), replace F(x) by σ0 W (x).
σ0 W (x)N (x) = σ0 D[W (x) u(x)] . (4.12)
Remove σ0 from both sides and integrate with respect to x .

W (x) N (x) dx = D[W (x) u(x)] dx . (4.13)
Work through the right-hand side of (4.13).

d[W (x) u(x)]
dx = d[W (x) u(x)] = W (x) u(x) − σ1 .
dx
Equate the left- and the right-hand sides of (4.13),

W (x) N (x) dx = W (x) u(x) − σ1 , (4.14)
transfer σ1 and divide the result by W (x). Voila !

We have the solution u(x) of the differential equation (4.3).

1
u(x) = W (x) N (x) dx + σ1 . (4.15)
W (x)
The unknown constant σ1 can be determined by a single boundary condition.

Note
Given below are examples group I and problems group I. They both retain the notation
M(x) and N (x) as in (4.2), and W (x) as in (4.10). The solution, u(x), is provided
by (4.15).
4.2 Examples Group I
Use integrating factor W (x) and solve the following twelve differential equations.
The notation is as in (4.2), (4.10), and (4.15).
3
(1) : M(x) = , N (x) = x 2 .
x
(2) : M(x) = 2 x , N (x) = x 3 .
3+x 2
(3) : M(x) = , N (x) = .
x x
1
(4) : M(x) = , N (x) = x 2 exp(x) .
x
(5) : M(x) = 3 tan(x) , N (x) = 2 sec(x) .
(6) : M(x) = 3 cot(x) , N (x) = 2 sin(2x) .
2x
(7) : M(x) = , N (x) = 5 exp(−x) .
x −2
2
(8) : M(x) = − , N (X ) = x 3 exp(−x) .
x
2
(9) : M(x) = − , N (x) = x − 1 .
x +1
2
(10) : M(x) = − , N (x) = (x − 1)2 .
x −1
1
(11) : M(x) = , N (x) = x 4 .
x log(x)
1
(12) : M(x) = + cot(x) , N (x) = cot(x) . (4.16)
x
4.2.1 Solution
Use (4.16)-(1)–(4.16)-(12): First, work out W (x) according to the procedure de-
scribed in (4.10). Next, use (4.15) to determine the solution
to the relevant differ-
ential equation. [Note: In problem (2) above, the integral x 3 exp(x 2 ) dx can be
4.2 Examples Group I 63
done as follows: Set x 2 = y. Then, 2x dx = dy and the integral can be written as

y exp(y) dy = 21 exp(y)(y − 1) = exp(x 2 ) (x 2−1) .]
2
1
2

3
(1) : W (x) = exp dx = exp log(x 3 ) = x 3 ;
x

1 1 x6
u = 3 x 3 × x 2 dx + σ0 = 3 + σ0 .
x x 6

(2) : W (x) = exp 2x dx = exp(x 2 ) ;

1 x2 − 1
u = x 3 exp(x 2 ) dx + σ0 = + σ0 exp(−x 2 ) .
exp(x 2 ) 2

3+x
(3) : W (x) = exp dx = exp[3 log(x) + x] = x 3 exp(x) ;
x

1 x 2 − 2x + 2 σ0
u = 3
2x 2 exp(x) dx + σ0 = 2 + 3 .
x exp(x) x3 x exp(x)

1
(4) : W (x) = exp dx = exp log(x) = x ;
x

1 1
u = x 3 exp(x) dx + σ0 = [exp(x)(x 3 − 3x 2 + 6x − 6) + σ0 ] .
x x

1
(5) : W (x) = exp 3 tan x dx = exp[−3 log(cos x)] = ;
(cos x)3

2
u = (cos x)3 sec x dx + σ0
(cos x)3
(sec x)3 (3 sin x + sin 3x)
= (cos x)3 (3 sin x + sin 3x) + σ0 = + σ0 (cos x)3 .
3 3

(6) : W (x) = exp 3 cot x dx = exp 3 log(sin x) = (sin x)3 ;

1 3 2 sin(2x) dx + σ
u = (sin x) 0
(sin x)3

1 10 sin x − 5 sin(3x) + sin(5x)
= + σ 0 .
(sin x)3 20

2x 2
(7) : W (x) = exp dx = exp 2 1+ dx
x −2 x −2

= exp 2x + 4 log(x − 2) = (x − 2)4 exp(2x) ;

1 4 exp(x) dx + σ
u = 5(x − 2) 0
(x − 2)4 exp(2x)
1
= 5(x 4 − 12x 3 + 60x 2 − 152x + 168) exp(x) + σ0 .
(x − 2) exp(2x)
4

−2 1 1
(8) : W (x) = exp dx = exp[−2 log(x)] = = 2 ;
x exp[log(x )]2 x

2 1 3 2
u = x x exp(−x) dx + σ0 = x − exp(−x) (x + 1) + σ0 .
x2

−2 1
(9) : W (x) = exp dx = exp −2 log(x + 1) = ;
x +1 (x + 1)2

(x − 1)
u = (x + 1)2 dx + σ0
(x + 1)2

2 2
= (x + 1) + log(1 + x) + σ0 .
(x + 1)

−2 1
(10) : W (x) = exp dx = exp[−2 log(x − 1)] = ;
x −1 (x − 1)2

(x − 1)2
u = (x − 1)2 dx + σ0 = (x − 1)2 [x + σ0 ] .
(x − 1)2

1 1
(11) : W (x) = exp dx = exp d(log x)
x log(x) log x

= exp log(log x) = log(x) ;
5
1 1 x x5
u = x 4 log(x) dx + σ0 = log(x) − + σ0 .
log(x) log(x) 5 25

1
(12) : W (x) = exp + cot x dx = exp log(x) + log(sin x)
x

= exp log(x sin x = x sin x ;

1 1
u = x sin x cot x dx + σ0 = x cos x dx + σ0
x sin x x sin x

1
= [x sin x + cos x + σ0 ] . (4.17)
x sin x
For given choices of the duo M(x) and N (x), solve the following differential equa-
tions labeled (1)–(12). [Hint: Compare (4.2), (4.10), and (4.15)–(4.17).]
1
(1) : M(x) = , N (x) = x .
x
3
(2) : M(x) = , N (x) = x 2 .
x
4.2 Examples Group I 65
2+x 3
(3) : M(x) = , N (x) = .
x x
1 + 3x
(4) : M(x) = , N (x) = x exp(x) .
x
(5) : M(x) = cot x , N (x) = sec x .
(6) : M(x) = tan x , N (x) = cos x .
x
(7) : M(x) = , N (x) = x exp(x) .
x −1
(8) : M(x) = 3 cot x , N (X ) = 2 cos x .
1
(9) : M(x) = − , N (x) = (x + 1)2 .
x +1
1
(10) : M(x) = − , N (x) = (x − 1)2 .
x −1
2
(11) : M(x) = , N (x) = x .
x log(x)
1
(12) : M(x) = + cot x , N (x) = cot x . (4.18)
x
4.3 Bernouilli Equation
When n = 0, the following equation
Du(x) + M(x)u(x) = N (x)u n (x) (4.19)
is similar to the first-order first-degree linear ordinary differential equation studied

in the preceding section. [see (4.2)]. Clearly, when n = 1, it is a first-order - first-
degree linear ordinary differential equation with constant coefficients. But it does
not outwardly appear to be so when n = 0 or 1. Yet it turns out that even when n is
different from 0 or 1, (4.19) can easily be reduced to linear form by a procedure first
suggested by Bernouilli.
4.3.1 The Bernouilli Suggestion
Introduce a function u 0 (x).
u 0 (x) = [u(x)]1−n . (4.20)
Therefore,
u(x) = [u 0 (x)]( 1−n ) ,

1
(4.21)
and

1 du 0 (x)
. [u 0 (x)]{( 1−n )−1} .
1
Du(x) = . (4.22)
1−n dx
Express u(x) and Du(x) as in (4.21) and represent (4.19) in terms of u 0 (x).

1 du 0 (x)
. [u 0 (x)]{( 1−n )−1} . + M(x) [u 0 (x)]( 1−n )
1 1
1−n dx
= N (x) [u 0 (x)]( 1−n )
n
. (4.23)
Multiply both sides by (1 − n) [u 0 (x)]−{( 1−n )−1} to arrive at the following linear
1
equation:
Du 0 (x) + (1 − n)M(x)u 0 (x) = (1 − n)N (x) [u 0 (x)]0

= (1 − n)N (x) . (4.24)
Bernouilli Equation: Solution

One can make the Bernouilli equation—that is (4.24)—look very similar to (4.2) by
introducing the notation
M0 (x) = (1 − n)M(x) ; N0 (x) = (1 − n)N (x) , (4.25)
and writing the resultant differential equation as
[D + M0 (x)]u 0 (x) = N0 (x) . (4.26)
Recall that a similar looking differential equation (4.2), that is,
[D + M(x)]u(x) = N (x) , (4.27)
has its solution embedded in (4.10) and (4.15). And both of these can immediately
be transferred to relate to the newest version of the Bernouilli differential equation—
that is (4.26)—merely by adding the subscript 0 to the functions u(x), M(x), and
N (x). So, according to (4.10) and (4.15), the solution to Bernouilli differential equa-
tion (4.26) is as follows:

1
u 0 (x) = W0 (x) N0 (x) dx + σ1 , (4.28)
W0 (x)
where, σ1 is an arbitrary constant and

4.3 Bernouilli Equation 67

W0 (x) = exp M0 (x) dx . (4.29)
With M0 (x) and N0 (x) as given in (4.25), and u 0 (x) as in (4.21), (4.28), and (4.29)
provide the desired solution of the differential equation (4.19).
4.3.2 Examples Group II
Make use of (4.20)–(4.29), and solve several Bernouilli equations of the form (4.19).
(1) Solve : [D + 2] u(x) = 3 [u(x)]−1 (4.30)
Equation (4.30): Solution

Here, M(x) = 2; N (x) = 3; and n = − 1. Therefore, (1 − n) = 2 and (1 −
n)M(x) = M0 (x) = 4, (1 − n)N (x) = N0 (x) = 6, and the resultant differential
equation, in the form of (4.26), is
[D + 4] u 0 (x) = 6 . (4.31)
According to (4.21), (4.28), and (4.29), we have

W0 (x) = exp M0 (x) dx

= exp 4 dx = exp (4 x) , (4.32)
and

1
u 0 (x) = W0 (x) N0 (x) dx + σ1
W0 (x)

= exp (−4 x) 6 exp (4 x) dx + σ1
6
= + σ1 exp (−4 x) = [u(x)]1−n = [u(x)]2 . (4.33)
4
Therefore, according to (4.33), the solution to the Bernouilli differential equa-
tion (4.30) is
21
6
u(x) = ± + σ1 exp (−4 x) . : (1) (4.34)
4
(2) Solve : [D + x] u(x) = x [u(x)]2 (4.35)


Here, M(x) = N (x) = x and n = 2. Therefore, M0 (x) = N0 (x) = − x,
(1 − n) = −1, and the resultant differential equation, in the form of (4.26), is
[D − x] u 0 (x) = − x (4.36)
According to (4.21), (4.28), and (4.29), one has

W0 (x) = exp M0 (x) dx
2
x
= exp − x dx = exp − , (4.37)
2
and

1
u 0 (x) = W0 (x) N0 (x) dx + σ1
W0 (x)
2 2
x x
= exp − x exp − dx + σ1
2 2
2
x
= 1 + σ1 exp = [u(x)]1−n = [u(x)]−1 . (4.38)
2
Therefore, the solution to the Bernouilli differential equation (4.35) is

2 −1
x
u(x) = 1 + σ1 exp . : (2) (4.39)
2
Next, solve
(3) : [D + x] u(x) = x [u(x)]3 (4.40)

Here, M(x) = N (x) = x and n = 3. Therefore, M0 (x) = N0 (x) = − 2 x
and (1 − n) = −2. The resultant differential equation, in the form of (4.26), is
[D − 2 x] u 0 (x) = − 2 x (4.41)
As usual, one has

W0 (x) = exp −2 x dx = exp −x 2 , (4.42)
and

u 0 (x) = exp x 2 (−2 x) exp −x 2 dx + σ1

= 1 + σ1 exp x 2 = [u(x)]1−n = [u(x)]−2 . (4.43)

− 21
u(x) = ± 1 + σ1 exp x 2 . : (3) (4.44)
Next, work out
(4) : [D + x] u(x) = exp(x 2 ) [u(x)]3 (4.45)

Here, M(x) = x; N (x) = exp(x 2 ); and n = 3. Therefore, (1 − n) = −2;
M0 (x) = − 2 x; N0 (x) = − 2 exp(x 2 ). The resultant differential equation, in
the form of (4.26), is
[D − 2 x] u 0 (x) = − 2 exp(x 2 ) (4.46)
Following the usual protocol, one gets

W0 (x) = exp −2 x dx = exp −x 2 , (4.47)
and

u 0 (x) = exp x 2 exp −x 2
. {−2 exp(x )} dx + σ1
2
= exp(x 2 ) [−2 x + σ1 ] = [u(x)]1−n = [u(x)]−2 . (4.48)
1
u(x) = ± . : (4) (4.49)
exp(x 2 ) [−2 x + σ1 ]
Another equation that to solve is

(5) : D + x 2 u(x) = exp(x 3 ) [u(x)]4 (4.50)

Here, M(x) = x 2 ; N (x) = exp(x 3 ); and n = 4. Therefore, (1 − n) = −3;
M0 (x) = − 3 x 2 ; N0 (x) = − 3 exp(x 3 ). The resultant differential equation, in
the form of (4.26), is

D − 3 x 2 u 0 (x) = − 3 exp(x 3 ) (4.51)
As usual, one can write

W0 (x) = exp −3 x 2 dx = exp −x 3 , (4.52)
and

u 0 (x) = exp x 3 exp −x 3 . {−3 exp(x 3 )} dx + σ1
= exp(x 3 ) [−3 x + σ1 ] = [u(x)]1−n = [u(x)]−3 . (4.53)
Therefore, the solution to the Bernouilli differential equation (4.50) is u(x) such that
1
[u(x)]3 = (4.54)
exp(x 3 ) [−3 x + σ1 ]
whose three solutions, according to standard rules of algebra, are
1
u(x) =
3
exp(x 3 ) [−3 x + σ1 ]
2
(−1) 3
u(x) =
3
exp(x 3 ) [−3 x + σ1 ]
4
(−1) 3
u(x) = . : (5) (4.55)
3
exp(x 3 ) [−3 x + σ1 ]
The next few Bernouilli type equations that are solved are numbered (6)–(10) and
are given below in (4.56).
(6) : [D + x] u(x) = x 3 [u(x)]−3

(7) : [D + (1/x)] u(x) = x [u(x)]2
(8) : [D + (1/x)] u(x) = x [u(x)]3
(9) : [D + (1/x)] u(x) = x [u(x)]−2

(10) : D + x 2 u(x) = exp(x) [u(x)]2 . (4.56)
Equations (4.56): Solution

(6): Here, M(x) = x; N (x) = x 3 ; and n = − 3. Therefore, (1 − n) = 4;
M0 (x) = 4 x; N0 (x) = 4 x 3. The resultant differential equation, in the form of
(4.26), is
[D + 4 x] u 0 (x) = 4 x 3 . (4.57)
As usual, we can write

W0 (x) = exp 4 x dx = exp 2 x 2 , (4.58)
and

u 0 (x) = exp −2 x 2 exp 2 x 2 . {4 x 3 } dx + σ1

1 2 2
= exp −2 x 2
exp 2 x (2x − 1) + σ1 . (4.59)
2
Therefore, the solution to the Bernouilli differential equation (4.56)-(6) is u(x) such
that

1
[u(x)]4 = (2x 2 − 1) + σ1 exp −2 x 2 (4.60)
2
whose four solutions, according to standard rules of algebra, are

1
1 4
u(x) = ± (2x − 1) + σ1 exp −2 x
2 2
2
1
1 4
u(x) = ±(i) (2x 2 − 1) + σ1 exp −2 x 2 . : (6) (4.61)
2
(7): Here, M(x) = 1/x; N (x) = x; and n = 2. Therefore, (1 − n) = −1;

M0 (x) = − 1/x; N0 (x) = − x. Thus,
[D − (1/x)] u 0 (x) = − x , (4.62)

dx 1
W0 (x) = exp − = exp (− log x) = , (4.63)
x x
and

u 0 (x) = x (−dx) + σ1
= [u(x)]−1 . (4.64)
Therefore, the solution to the Bernouilli differential equation (4.56)-(7) is
1
u(x) = . (4.65)
x [−x + σ1 ]
(8): Here, M(x) = 1/x; N (x) = x; and n = 3. Therefore, (1 − n) = −2;

M0 (x) = − 2/x; N0 (x) = − 2 x, and
[D − (2/x)] u 0 (x) = − 2 x , (4.66)

dx 1
W0 (x) = exp −2 = exp (−2 log x) = , (4.67)
x x2

dx
u 0 (x) = x −2
2
+ σ1
x
= [u(x)]−2 . (4.68)
Accordingly, the solution to the Bernouilli differential equation (4.56)-(8) is

− 1
u(x) = ± x 2 {−2 log(x) + σ1 } 2 . (4.69)
(9): Here, M(x) = 1/x; N (x) = x; and n = − 2. Therefore, (1 − n) = 3;

M0 (x) = 3/x; N0 (x) = 3 x. Thus,
[D + (3/x)] u 0 (x) = 3 x , (4.70)

dx
W0 (x) = exp 3 = exp (3 log x) = x 3 , (4.71)
x
and

u 0 (x) = x −3 3 x 4 dx + σ1

3
= x −3 x 5 + σ1 = [u(x)]3 . (4.72)
5
Therefore, there are three solutions to the Bernouilli differential equation (4.56)-(9):
namely
3 5 1
x + σ1 3
u(x) = 5
,
x3
3 5 1
2 x + σ 1
3
u(x) = (−1) 3 5
,
x3
3 5 1
4 x + σ1 3
u(x) = (−1) 3 5
. : (9) (4.73)
x3
(10): Here, M(x) = x 2 ; N (x) = exp(x); and n = 2. Therefore,

(1 − n) = −1; M0 (x) = − x 2 ; N0 (x) = − exp(x). Thus,

D − x 2 u 0 (x) = − exp(x) , (4.74)
3
x
W0 (x) = exp −x 2 dx = exp − , (4.75)
3
and

x3 x3
u 0 (x) = exp − exp x − dx + σ1 . (4.76)
3 3
The solution to the Bernouilli differential equation (4.56)-(10) is
u(x) = [u 0 (x)]−1 . (4.77)
Given below are a set of ten Bernouilli differential equations similar to (4.19). The
relevant choices for n, M(x) and N (x), are as noted in (4.78)-(1)–(4.78)-(10). [Hint:
Compare (4.19)–(4.77).]
(1) : M(x) = 3 , N (x) = 2 , n = − 1 .

(2) : M(x) = 2 x , N (x) = 3 x , n = 2 .
(3) : M(x) = 3 x , N (x) = 2 x , n = 3 .
(4) : M(x) = x , N (x) = 3 exp(x 2 ) , n = 3 .
(5) : M(x) = x 2 , N (x) = 6 exp(x 3 ) , n = 4 .
(6) : M(x) = x , N (x) = 2 x 3 , n = − 3 .
1
(7) : M(x) = , N (x) = 3 x , n = 2 .
x
2
(8) : M(x) = , N (X ) = 2 x , n = 2 .
x
2
(9) : M(x) = , N (X ) = 2 x , n = 3 .
x
2
(10) : M(x) = , N (X ) = 3 x , n = − 2 . (4.78)
x
Chapter 5
Green’s Function Laplace Transforms
5.1 Green’s Function
Consider an inhomogeneous linear ordinary differential equation.
V (x) Y (x) = F(x) . (5.1)
The differential operator V (x), the solution Y (x), and quite possibly also the
inhomogeneous term represented by the arbitrary function F(x), involve constants
and derivatives with respect to the variable x. The objective of the present exercise
is to determine the solution, Y (x), of the differential equation (5.1) with specified
boundary conditions.
In the preceding chapters, various straightforward methods for solving homo-
geneous linear ordinary differential equations were discussed. For inhomogeneous
linear ordinary differential equations, those methods required more effort. Espe-
cially, the particular integral, Ipi , needs to be worked out ab initio for every different
inhomogeneous term F(x). But there exists another powerful methodology named
after George Green26. whereby Green’s function, once and for all, helps solve the
differential equation for arbitrary values of the inhomogeneous term. Therefore, for a
particular differential operator, in addition to the homogeneous solution, only Green’s
function needs to be worked out.
The Green function procedure is well-suited to studying inhomogeneous ordinary
differential equations. A given Green’s function always refers to a particular differ-
ential operator. For the present case, V (x) is the relevant differential operator. While
there is no total agreement on this issue, most of the available literature, for instance,
Dean J. Duffy24. , would accept the following definition of Green’s function, G(x, x ),
that is, relevant to the differential operator V (x).
V (x) G(x, x ) = δ(x − x ) . (5.2)

https://doi.org/10.1007/978-3-319-76406-1_5
76 5 Green’s Function Laplace Transforms
Here, δ(x − x ) is Dirac’s delta function. Note, some discussion of the Dirac33. delta
function is provided in the Appendix.
If Green’s function, G(x, x ), as defined in (5.2), can be worked out, the desired
solution of the differential equation V (x) Y (x) = F(x) can be represented as (5.3)
given below.

Y (x) = F(x ) G(x, x ) dx . (5.3)
The correctness of this assertion is assured if it can be shown to be consistent with the
fundamental equation (5.1). To check this fact, let us proceed as follows. Multiply
(5.3) from the left by V (x).

V (x) Y (x) = V (x) F(x ) G(x, x ) dx

= F(x ) V (x) G(x, x ) dx . (5.4)
And use (5.2).

V (x) Y (x) = V (x) F(x ) G(x, x ) dx

= F(x )V (x) G(x, x ) dx = F(x ) δ(x − x ) dx = F(x) . (5.5)
Notice that (5.5) properly reproduced (5.1).

Let us work with boundary conditions that apply at the ends of the allowed physical
space. Assume that such physical space ranges over the interval 0 ≤ x ≤ π—and
choose the following as the boundary conditions:
Y (x = 0) ≡ Y (0) = 0 , Y (x = π) ≡ Y (π) = 0 . (5.6)
Because (5.3) holds for arbitrary F(x ), these boundary conditions,

π
Y (0) = F(x ) G(0, x ) dx = 0 ,
0
π
Y (π) = F(x ) G(π, x ) dx = 0 , (5.7)
0
demand
G(0, x ) = 0 , G(π, x ) = 0 . (5.8)
An important property of G(x, x ) is that (5.3) applies to any inhomogeneous

term F(x ) subject only to the requirement that F(x ) is well behaved. Clearly,
5.1 Green’s Function 77
if one can determine Green’s function, the relevant inhomogeneous differential

equation (5.1) is readily solved and (5.3) provides the desired solution Y (x) for
arbitrary choice of the inhomogeneous term F(x ). The question then is to learn how
to travel the route that leads to the needed Green’s function.
There are several possible routes: some more tortuous than others. And while
the reached Green’s function may behave as expected, the chances are that at least
occasionally it would lead to solutions that are too complicated to be useful.
5.2 Solving Differential Equations
5.2.1 Eigenfunction Expansion
Most differential equations have two components: One that deals with the homo-
geneous part of the equation and the other that involves the inhomogeneous part.
Fortunately, the various techniques for handling the homogeneous part are similar
to those discussed in detail in Chaps. 3 and 4 . Therefore, it can be assumed that the
solution to the homogeneous part can be worked out.
In this subsection, we show that a Green’s function may be calculated by using
eigenfunctions. That is, information obtained from solving the homogeneous part of
the differential equation. Indeed, a Green’s function is often fathered by the homo-
geneous part of the referencing differential equation. And once Green’s function is
calculated, it can be used to solve the inhomogeneous part of the referencing differ-
ential equation. And that can be done for arbitrary choices of the inhomogeneous
term.
Let us consider the following linear ordinary differential equation.
2
D + α2 yn (x) = γn yn (x) , n = 0, 1 , 2 , 3 , . . . (5.9)
This equation is similar to differential equations we have learned to solve in

Chaps. 3, and 4. Equation
(5.9) is actually an eigenvalue equation34. . The Hamil-
tonian, D + α , is Hermitian and the eigenfunctions, yn (x),
2 2
yn (x) = A sin [n x] + B cos [n x] , (5.10)
are orthogonal and complete. Also the eigenvalues, γn , are real. As shown below,
for (5.9) both the eigenvalues γn and the eigenfunction yn (x) are readily evaluated.
The eigenfunction can also be made orthonormal so that the relevant integral of
eigenfunctions yn1 (x) and yn2 (x), calculated over the prescribed spatial range, equals
δn1 , n2 .
π
yn1 (x) yn2 (x) dx = δn1 , n2 . (5.11)
0
We require that the eigenfunctions yn (x) remain valid in the physical interval
0 ≤ x ≤ π and obey the boundary conditions
yn (0) = 0 , yn (π) = 0 . (5.12)
The first boundary condition, namely yn (x = 0) = 0, is satisfied only by the sine

term in (5.10). And because at x = 0 the cosine term is non-vanishing, cosines must
be excluded altogether from (5.10). Similarly, regarding the sign term, the second
boundary condition—namely yn (x = π) = 0—requires that for other than the trivial
case n = 0, sin [n π] is equal to zero only when n is an integer. Therefore,
yn (x) = |ρn | sin(n x) , for n = 1, 2, 3, . . . (5.13)
Without loss of generality, one can use only the positive n. The normalization require-
ment,
π π
π
[yn (x)] [yn (x)]∗ dx = 1 = |ρn |2 sin(n x)2 dx = |ρn |2 , (5.14)
0 0 2
sets the constant

2
|ρn | = (5.15)
π
and the eigenfunctions yn (x).

2
yn (x) = sin(n x) . (5.16)
π
Using (5.9) and (5.16), we can write
[D2 + α2 ] yn (x) = γn yn (x)

2 2
= [D + α ]
2 2
sin(n x) = [α − n ]
2 2
sin(n x) = [α2 − n2 ] yn (x) .
π π
(5.17)
Equation (5.17) provides the relevant eigenvalues γn .
γn = (α2 − n2 ). (5.18)
5.2 Solving Differential Equations 79
5.2.2 Green’s Function Calculated
As stated above, the eigenfunctions yn (x) are orthonormal and complete. [Note : For
observational convenience, (5.1) is reprinted below as (5.19).] Therefore, when Y (x)
F(x) = V (x) Y (x) (5.19)
is a function in the same domain as the eigenfunctions yn (x) and obeys the same
boundary conditions, it may be represented as a linear combination of the stated
eigenfunctions.
∞

Y (x) = bn yn (x). (5.20)
n=0
To study this matter, further proceed as follows:

Set V (x) = [D2 + α2 ]. Reprint (5.2). And insert the Y (x), given in above (5.20),
into (5.19).
V (x) = [D2 + α2 ] ;
V (x) G(x, x ) = δ(x − x ) ;
∞ ∞

F(x) = [D + α ]
2 2
bn yn (x) = bn [D2 + α2 ] yn (x) . (5.21)
n=0 n=0
Use the first row of (5.17) and thereby replace [D2 + α2 ] yn (x) by γn yn (x). As a
result, (5.21) becomes
∞

F(x) = bn γn yn (x) . (5.22)
n=0
The constant bn needs to be determined. To that end, multiply (5.22) by ym (x) and
integrate over the relevant space.
π π ∞

ym (x) F(x) dx = bn γn ym (x) yn (x)dx
0 0 n=0
∞
π ∞

= bn γn ym (x) yn (x)dx = bn γn δn,m = bm γm . (5.23)
n=0 0 n=0
Now divide both sides of (5.23) by γm .

π
1
ym (x) F(x) dx = bm . (5.24)
γm 0
For convenience, change the variables from m to n and x to x in (5.24). This gives
us the desired bn .
π
1
bn = yn (x ) F(x ) dx . (5.25)
γn 0
Combine the result in (5.25) for bn with (5.20) to get

∞
∞ π
1
Y (x) = bn . yn (x) = yn (x ) F(x ) dx . yn (x)
n=0 n=1
γ n 0
π ∞

1
= yn (x) yn (x ) F(x ) dx .

(5.26)
0 n=0 n
γ
But according to (5.3), we have

π
Y (x) = G(x, x ) F(x ) dx . (5.27)
0
The eigenfunction Green’s function G(x, x ) is now readily identified by treating

simultaneously (5.26) and (5.27). We get
∞ ∞
1 1
G(x, x ) = yn (x) yn (x ) = y (x) yn (x ).
2) n
(5.28)
n=0
γn n=0
(α 2 − n
[Note: The eigenvalue γn used in (5.28) is picked up from (5.18).]

It should be emphasizedthat the denominator
(α2 − n2 ) in (5.28) is particular
to the differential operator D + α . And according to (5.27), for any arbitrary
2 2
choice of F(x ), Green’s function (5.28) should provide the desired solution Y (x).
In other words, the function Y (x) in (5.29),

D2 + α2 Y (x) = F(x) , (5.29)
is determined by the relationship

∞
π
1
Y (x) = yn (x) yn (x ) F(x ) dx
n=0
(α − n2 )
2
0
∞ π
2 1
= sin(n x) sin(n x ) F(x ) dx . (5.30)
π n=0 (α2 − n2 ) 0
5.2.3 Examples Group I
Use (5.16) and (5.30) and work out solutions to the differential equation D2 + α2 Y (x) = F(x)
for the following choices of F(x).
F(x) = sin(x) . (1)

F(x) = sin(x) cos(x) . (2)
F(x) = x . (3) (5.31)
5.2.4 Solution
(1)
Insert F(x ) = sin(x ) in (5.30).
∞ π
2 1
Y (x) = sin(n x) sin(n x ) sin(x ) dx
π n=0 (α2 − n2 ) 0
π
Because the integral 0 sin(n x ) sin(x ) dx = ( π2 )δ(n − 1), only the n = 1 term sur-
vives. Thus,
∞
2 1 π sin(x)
(1) : Y (x) = sin(x) δ(n − 1) = 2 . (5.32)
n=0
π (α2 −n )
2 2 α −1
(2)
sin(2 x )
Insert F(x ) = sin(x )cos(x ) = 2
in (5.30).
∞ π
1 2 1
Y (x) = sin(n x) sin(n x ) sin(2 x ) dx
2 π n=0 (α2 − n2 ) 0
π
Because the integral 0 sin(n x ) sin(2 x ) dx = ( π2 )δ(n − 2), only the n = 2 term
survives. Thus,
∞
1 1 π 1 sin(2x)
(2) : Y (x) = sin(n x) δ(n − 2) = .
n=0
π (α2 − n2 ) 2 2 α2 − 4
(5.33)
(3)
Insert F(x ) = x in (5.30).
∞ π
2 1
: Y (x) = sin(n x) sin(n x ) x dx (5.34)
π n=0 (α2 − n2 ) 0
π
Do the integral 0 sin(n x ) x dx by parts.

π
x cos(n x) π 1 π
sin(n x) x dx = − + cos(n x)dx
0 n 0 n 0

π cos(n π) 1 sin(nx) π π(−1)n+1
=− + = +0 . (5.35)
n n n 0 n
In order to calculate Y (x), the result in (5.35) above should be inserted into (5.34).
∞
(−1)n+1
(3) : Y (x) = 2 sin(n x) . (5.36)
n=0
n(α2 − n2 )
Note: The result Y (x) given in (5.36) looks more complicated than one would have
anticipated. Therefore, it is important to check its accuracy. To that end let us plug
this Y (x) into (5.29) and check whether it actually leads to F(x) = x. That is, let us
work with equations
2
D + α2 Y (x) = F(x)
∞
(−1)n+1
= D 2 + α2 2 sin(n x) = F(x) = x
n=0
n(α2 − n2 )
∞
∞
(−1)n+1 (−1)n+1
= 2 (α2 − n2 ) sin(n x) = 2 sin(n x) . (5.37)
n=0
n(α − n )
2 2
n=0
n
Again one wonders whether the right-hand side of (5.37) is indeed equal to x. Actu-
ally, it looks much like a Fourier series. If so, is it the Fourier expansion of x? Let us
check.
Fourier expansion of x.
∞
∞

x = a0 + an cos(n x) + bn sin(nx) , (5.38)
n=1 n=1
where
π
1 1 π
a0 = x dx = 0 , an = x cos(nx)dx = 0 ,
2 π −π π −π

1 π 1 −x cos(n x) π 1 sin(n x) π
bn = x sin(nx)dx = +
π −π π n −π π n −π
n+1
1 2 π (−1) (−1) n+1
= = 2 . (5.39)
π n n
Using bn from (5.39), (5.38) gives

∞
(−1)n+1
x = 2 sin(n x) . (5.40)
n=0
n
Comparison of (5.37) and (5.40) confirms the result F(x) = x. Q.E.D.

Results for Differential Operator V (x) = D2 − 9
4
Set a numerical value for the variable α.

3
α= . (5.41)
2
As a result, the previously calculated expression for the eigenfunction Green’s func-
tion given in (5.28), i.e.,
∞
1
G(x, x ) =eigenfunction result yn (x) yn (x ) ,
n=0
(α2 − n2 )
will change. And we will have the following eigenfunction Green’s function when
α = 23 .
∞
4
G(x, x ) = eigenfunction result yn (x) yn (x ). (5.42)
n=0
(9 − 4 n2 )
5.3 Calculation by Approaching Delta Function
As noticed above, the eigenfunctions Green’s function leads to infinite series with
results that are often both difficult to work out and complicated to work with. On
many occasions, a better approach is to work directly with the defining equation of
Green’s function that involves Dirac’s delta function. And then approach the delta
function singularity from either sides, thereby obtaining a closed-form expression for
Green’s function. To demonstrate this procedure and to compare with the previous
results, we work with a similar differential equation to that used for deriving the
eigenfunctions Green’s function.
Consider the differential operator V (x) = D2 + 49 and its Green’s function
G(x, x0 ). [Note: Compare (5.2)]

9
D2 + G(x, x0 ) = δ(x − x0 ) . (5.43)
4
As before, x and x0 are chosen to lie within the interval 0 ≤ x ≤ π. The delta func-
tion in (5.43) refers to the physical separation (x − x0 ). There are two possibilities
for reaching the singularity at x = x0 . First: x can approach x0 from below. Second:
x approaches x0 from above. For either of these options, as long x dos not touch x0 ,
the delta function δ(x − x0 ) itself is vanishing and (5.43) reduces to the following:

9
D +
2
G(x, x0 ) = 0 , (5.44)
4
with the solution
3 3
G(x, x0 ) = A sin x + B cos x . (5.45)
2 2
Let us examine how the constants A and B are affected as the position x moves within
the intervals x0 > x ≥ 0 and π ≥ x > x0 .
One of the relevant boundary conditions (5.8), namely
G(x = 0, x0 ) = 0 , (5.46)
excludes the cosines in the region x0 > x ≥ 0. Therefore, we have
3
G(x, x0 ) = A sin x , x0 > x ≥ 0. (5.47)
2
In contrast, for the region π ≥ x > x0 we need to have, G(x = π, x0 ) = 0. Therefore,

the sine term is excluded. As such, we can write
3
G(x, x0 ) = B cos x , π ≥ x > x0 . (5.48)
2
Our next task is to calculate the constants A and B. To that end, we can impose the
continuity requirement at x0 so the result is the same whether x0 is approached from
below or from above. We get
3 3
A sin x0 = B cos x0 . (5.49)
2 2
Clearly, we need one more relationship

to determine A and B, that is, provided by
0)
continuity of the differential dG(x,x
dx
. This is best calculated by integrating
x=x0
(5.43) in the following manner
5.3 Calculation by Approaching Delta Function 85

x0 + x0 +
9
lim→o D2 + G(x, x0 ) dx = lim→o δ(x − x0 )
x0 − 4 x0 −
x0 + x0 +
dG 9
= lim→o + lim→o G(x, x )dx = 1 . (5.50)
dx x0 − 4 x0 −
Equation (5.50) leads to

dG dG
lim→o − +0 = 1. (5.51)
dx x0 + dx x0 −
Using the expressions (5.48) and (5.47), (5.51) is rewritten as
3 3 3 3
−B sin x0 − A cos x0 =1 . (5.52)
2 2 2 2
Solving (5.49) and (5.52) together gives
2 3 2 3
A = − cos x0 ; B = − sin x0 . (5.53)
3 2 3 2
Inserting A and B given in (5.53) into (5.47) and (5.48) leads to the desired Green’s
function.
2 3 3
G(x, x0 ) = − cos x0 sin x , x0 > x ≥ 0. (5.54)
3 2 2
2 3 3
G(x, x0 ) = − sin x0 cos x , π ≥ x > x0 . (5.55)
3 2 2
Use the closed-form Green’s function procedure outlined in (5.54) and

(5.55) and work out the differential equation (5.29) for the same two choices
of F(x) given in (5.31) that were treated with the eigenfunctions Green’s func-
tion D2 + α2 Y (x) = F(x).
F(x) = sin(x) . (1)

F(x) = sin(x) cos(x) . (2) . (5.56)
5.3.2 Solution
F(x) = sin(x) : (1)

Rewrite (5.27).
π
Y (x) = G(x, x0 ) F(x0 ) dx0 .
0
Exchange variables x and xo .

π
Y (x0 ) = G(x0 , x) F(x) dx .
0
Because
the current
Green’s function is produced by a self-adjoint differential oper-
ator, D2 + α2 , its Green’s function is symmetric, i.e., G(x0 , x) = G(x, x0 ). There-
fore, we can write the above as
π
Y (x0 ) = G(x, x0 ) F(x) dx . (5.57)
0
Insert the G(x, x0 ) given in (5.54) and (5.55) into (5.57). For F(x) = sin(x), we get
x0
2 3 3
Y (x0 ) = − cos x0 sin x sin(x) dx
3 2 0 2
π
2 3 3
− sin x0 x sin(x) dxcos
3 2 x0 2

2 3 x0 1 5
=− cos x0 sin − sin x0
3 2 2 5 2
x
2 3 0 1 5
− sin x0 −cos + cos x0 . (5.58)
3 2 2 5 2
Equation (5.58) can be further organized.

x
2 x0 3 0 3
Y (x0 ) = − sin cos x0 − cos sin x0
3 2 2 2 2

2 5 3 5 3
+ sin x0 cos x0 − cos x0 sin x0
15 2 2 2 2

2 x0 x0 2 5 x0 3 x0
=− sin −3 + sin −
3 2 2 15 2 2
= (2/3) sin(x0 ) + (2/15) sin(x0 ) = (4/5) sin(x0 ) . (5.59)
It is straightforward to check the corresponding prediction of the eigenfunctions

Green’s function. For that purpose, we go to (5.32) and record its result obtained
from summing an infinite series. That is
sin(x)
(1) : Y (x) = . (5.60)
α2 − 1
The only change necessary in the eigenfunctions result given in (5.60) is to change
α2 to 94 . Then, we would have Y (x) = 4 sin(x)
5
, which is exactly the same as the current
result given above in (5.59). Q.E.D.
F(x) = sin(x) cos(x) : (2)
In order to calculate Y (x0 ) with F(x) = sin(x) cos(x), we need to insert Green’s
function provided in (5.54) and (5.55) into (5.57). We get
x0
1 3 3
Y (x0 ) = − cos x0 sin x sin(2 x) dx
3 2 0 2
π
1 3 3
− sin x0 cos x sin(2 x) dx . (5.61)
3 2 x0 2
After doing the integrals, the above can be written as

1 3 x0 1 7 x0
Y (x0 ) = − cos x0 sin − sin
3 2 2 7 2
x
1 3 0 1 7 x0
− sin x0 cos + cos
3 2 2 7 2
x
1 x0 3 x0 3 x0 0
=− sin cos + sin cos
3 2 2 2 2

1 7 x0 3 x0 3 x0 7 x0
+ sin cos − sin cos
21 2 2 2 2
1 x0 3 x0 1 7 x0 3 x0
=− sin + + sin −
3 2 2 21 2 2
1 1
=− sin (2 x0 ) + sin (2 x0 )
3 21
2
=− sin(2 x0 ) . (5.62)
7
Again, it is straightforward to check the corresponding prediction of the eigen-

functions Green’s function. For that purpose, we go to (5.33) and record its result
obtained from summing an infinite series. That is,
1 sin(2x)
(2) : Y (x) = . (5.63)
2 α2 − 4
The only change necessary in the eigenfunctions result given in (5.63) is to change
α2 to 49 . Then, we would have Y (x) = − 27 sin(2 x0 ), which is exactly the current
result given above in (5.62). Q.E.D.
Work out Green’s function relevant to differential equation (5.64).

d
− α x(t) = t 2 exp(ω t) . (5.64)
dt
5.3.4 Solution
Similar to (5.1), (5.2), (5.3) we can write
V (t) Y (t) = F(t) , (5.65)

V (t) G(t, t0 ) = δ(t − t0 ) , (5.66)
π
Y (t) = F(t0 ) G(t, t0 ) dt0 . (5.67)
0
Next, we choose

d
V (t) = − α , Y (t) = x(t) , F(t) = t 2 exp(ω t) , (5.68)
dt
work out the relevant Green’s function, G(t, t0 ), determine the solution x(t) of the
differential equation (5.69) within the domain {π ≥ t > 0} and obeying the boundary
condition {limt=0+ x(t) = 0 .}

d
− α x(t) = t 2 exp(ω t) . (5.69)
dt
In order to follow the instructions given above, begin with (5.66).


d
− α G(t, t0 ) = δ(t − t0 ) . (5.70)
dt
For t > t0 , the delta function δ(t − t0 ) is vanishing. Then, the truncated version of
differential equation (5.70) is easy to solve and we get
G(t, t0 ) = exp [α (t − t0 )] , for t > t0 . (5.71)
Therefore, according to (5.67) and (5.68) the solution x(t) is

π
x(t) ≡ Y (t) = F(t0 ) G(t, t0 ) dt0 . (5.72)
0
Set F(t0 ) = t02 exp(ω t0 ). Insert it in (5.72). Use (5.71) and rewrite the result as (5.73).
π t
x(t) = t02 exp(ω t0 ) G(t, t0 ) dt0 = t02 exp(ω t0 ) exp [α (t − t0 )] dt0
0 0
t
= exp (α t) t02 exp [t0 (ω − α)] dt0

0
t t
t02 exp [t0 (ω − α)] 2
= exp (α t) − t0 exp [t0 (ω − α)] dt0
(ω − α) 0 ω−α 0

t 2 exp ω t 2 exp α t t0 exp [t0 (ω − α)] t
= −
ω−α ω−α (ω − α) 0
t
2 exp α t
+ exp [t0 (ω − α)] dt0
(ω − α)2 0

t2 2t 2 2 exp α t
= (exp ω t) − + − . (5.73)
(ω − α) (ω − α) 2 (ω − α) 3 (ω − α)3
In order to ascertain whether the x(t) given in (5.73) satisfies the differential
equation (5.69), we proceed as follows.

dx(t) 2t 2 2 α exp α t
− α x = (exp ω t) − −
dt (ω − α) (ω − α)2 (ω − α)3
2
t 2t 2
+ ω (exp ω t) − +
(ω − α) (ω − α)2 (ω − α)3

t2 2t 2 2 α exp α t
− α (exp ω t) − + + .
(ω − α) (ω − α) 2 (ω − α) 3 (ω − α)3
(5.74)
Equation (5.74) meanders and is quite long. We write it in a more organized form
below.

dx(t) ω−α 2α − 2 α
− α x(t) = t 2 exp(ω t) + exp(α t)
dt ω−α (ω − α)3

(2 ω − 2 α) 2
+ exp(ω t) − = t 2 exp(ω t) . (5.75)
(ω − α) 3 (ω − α)2
This is indeed the result expected from the differential equation (5.69). Therefore,
the correctness of the x(t) given in (5.73) is verified. Q.E.D.
5.4 Laplace Transform
It is helpful to use a general notation for Laplace transforms. Consider some function
j of t, as in j(t), where j is any lowercase character. Denote its Laplace transform by
the same uppercase character J . But signify it as a function of s, as in J (s). As such
the inverse transform of J (s) is j(t). These statements are displayed below in (5.76).
t0
s {j(t)} = limt0 −>∞ j(t) exp(−s t) dt ≡ J (s),
0
−1
s {J (s)} ≡ j(t) . (5.76)
For a function j(t) such that j(t) = 0 when t < 0, the Laplace integral s {j(t)}
specifies the Laplace transform J (s).
∞
J (s) ≡ s {j(t)} = j(t) exp(−s t) dt. (5.77)
0
J (s) exists for s > 0 provided j(t) satisfies the following conditions:
(1) j(t)= 0 for t < 0
(2) j(t) is continuous, or at least piecewise continuous, in every interval
(3) t n f (t) < ∞ as t− > 0 for some number n, where n < 1.
5.4 Laplace Transform 91
5.4.1 Table

f (t) F(s) =limt0 −>∞
t0
f (t) exp(−s t) dt
0

(1). 1 1/s

(2). a1 f1 (t) + a2 f2 (t) a1 F1 (s) + a2 F2 (s)

(3). H (t − c) exp(−c s)/s

(4). H (t − c)f (t − c) exp(c s)f (s)

(5). t 1/s2

(6). t 2 3
2!/s
(7). t 3 4
3!/s
(8). t n , n = 1, 2, 3 n!/sn+1

(9). t p , p > −1 G(p+1)

sp+1
(10). exp(a t) , s > a 1/(s − a)

(11). t exp(a t) , s > a 1/(s − a) 2

(12). δ(t − a) exp(−a s)

(13). t n exp(a t) , s > a n!/(s − a)n+1

(14). cos(b t) s/(s + b )
2 2

(15). exp(a t)cos(b t) (s − a)/[(s − a)2 + b2 ]

(16). cosh(b t) s/(s − b )
2 2

(17). sin(b t) b/(s2 + b2 )

(18). exp(a t)sin(b t) b/[(s − a)2 + b2 ]

(19). sinh(b t) b/(s − b )
2 2

(20). t sin(b t) (2 a s)/(s + b )
2 2 2

(21). t cos(b t) (s2 − b2 )/(s2 + b2 )2

: (22). t sinh(a t) 2 a s/(s − a )
2 2 2
(5.78)

(23). t cosh(a t) s2 + a2 (s2 − a2 )2

(24). sin(a t) − a t cos(a t) 2 a /(s + a )
3 2 2 2

(25). sin(a t) + a t cos(a t) 2 a s2 /(s2 + a2 )2

(26). cos(a t) − a t sin(a t) s (s2 − a2 )/(s2 + a2 )2

(27). cos(a t) + a t sin(a t) s (s + 3 a )/(s + a )
2 2 2 2 2

(28). sin(a t + b) [s sin(b) + a cos(b)]/(s + a ) 2 2

(29). cos(a t + b) [s cos(b) − a sin(b)]/(s2 + a2 )

(30). 1 f (t) ∞
t F(u) du
(31). ft (c t) 1 s
F
t
c c

(32). 0 f (v) dv 1
s F(s)

(33). t f (t − τ ) g(τ ) dτ
F(s) G(s)
0
(34). cos(a t + b) [s cos(b) − a sin(b)]/(s2 + a2 )

(35). f (t) s F(s) − f (0)

(36). f (t)
s F(s) − s f (0) − sf (0) − f (0)
3 2

(37). [exp(a(a−b) t)−exp(b t)]
(s − a)−1 (s − b)−1

√
(38). erfc √ a 1
exp −a s
2 t s
√
(39). √1 exp −a2 √1 exp(−a s)
πt 4 t s
√
(40). √a exp −a2 exp(−a s)
π 4t

2 t 3
−1 −1
(41). a exp(a t)−b exp(b t)
s(s − a) (s − b)
a−b
π
(42). √t
4s
5.5 Computation
Laplace transform of a function f (t)—which is at least piecewise continuous on

a specified interval—is defined through the use of an integral and is informatively
denoted as s { f (t)}. In other words,
t0
s { f (t)} = limt0 −>∞ f (t) exp(−s t) dt = F(s). (5.79)
0
To get a feel for how the integration process (5.79) actually unfolds, let us work
out a few simple examples.
(1) : For f (t) = t n , compute s { f (t)} = F(s).
To solve (1) write:
t0 n ∞

s t n = limt −>∞ t n exp(−s t) dt ≡ t (n−1) exp(−s t) dt
0
0 s 0
∞
n(n − 1)
= t (n−2) exp(−s t) dt
s2 0
∞
n(n − 1)(n − 2)
= t (n−3) exp(−s t) dt
s3 0
= ...
n n!
Therefore , s t = . (5.80)
s(n+1)
In particular, using n = 0, and n = 1, we get s {1} = 1s , and s {t} = 1

s2
.
(2) : For f (t) = sin(b t), compute s { f (t)} = F(s).
To solve (2) write:
f (t) = sin(b t) , for t > 0 ,

t0
F(s) = limt0 −>∞ sin(b t) exp(−s t) dt , s > 0 ,
0
Doing integration by parts gives:

cos(b t) exp(−s t) t0 s t0
= limt0 −>∞ − − cos(b t) exp(−s t) dt ,
b 0 b 0
2 t0
1 s
= − limt0 −>∞ sin(b t) exp(−s t) dt ,
b b2 0
1 s2
F(s) = − F(s) . (5.81)
b b2
5.5 Computation 93
Therefore,
b
s {sin(b t)} = F(s) = . (5.82)
s2 + b2
t
NOTE: For convenience, instead of the proper form, i.e., limt0 −>∞ 0 0 , of the
infinite integrals that occur here, heretofore we shall use only their improper
∞
form, i.e., 0 .
(3) : For f (t) = c1 f1 (t) + c2 f2 (t) compute s { f (t)} .
To solve (3) write:
∞
s {c1 f1 (t) + c2 f2 (t)} = f (t) exp(−s t) dt
0
∞ ∞
= c1 f1 (t) exp(−s t) dt + c2 f2 (t) exp(−s t) dt
0 0
= c1 s { f1 (t)} + c2 s { f2 (t)}
= c1 F1 (s) + c2 F2 (s). (5.83)
(4) : For f (t) = exp(at), compute s { f (t)} = F(s).

To solve (4) write:
∞
s {exp(at)} = exp[−(s − a) t] dt ,
0
1
= , only for s > a . (5.84)
s−a
Here, it is convenient to briefly describe H (t − a), the Heaviside step function.

However, for greater detail see (5.92). The step function H (t − a) is defined for
a ≥ 0 by the relationship
H (t − a) = 1, for t > a
= 0, for t < a . (5.85)
(5) : For f (t) = f1 (t − a) H (t − a), compute s { f1 (t − a) H (t − a)} = F(s).

To solve (5) write:
∞
s { f1 (t − a) H (t − a)} = f1 (t − a) H (t − a) exp(−s t)dt
0 ∞
= f1 (t − a) exp(−s t) dt .
a
Set t − a = y : then dt = dy and the above becomes

∞
s { f1 (t − a) H (t − a)} = f1 (y) exp[−s (y + a)] dy
0
∞
= exp(−s a) f1 (t) exp[−s t] dt
0
= exp(−s a) s { f1 (t)} . (5.86)
(6) : Consider f (t) = δ(t − a). Note that a ≥ 0.

To solve (6) write:
∞
s {δ(t − a)} = exp(−s t)δ(t − a) dt
0
= exp(−s a) . (5.87)
dfi (t) d2 fi (t) d3 fi (t)

(7) : First consider f (t) = dt
, f (t) = dt @
, f (t) = dt 3
.
To solve (7) write:
∞
d3 fi (t) d3 fi (t)
s = exp(−s t) dt
dt 3 0 dt 3
∞
d2 fi (t) ∞ d2 fi (t)
= exp(−s t) |0 + s exp(−s t) dt.
dt 2 0 dt 2
∞
d2 fi (t) dfi (t) ∞ dfi (t)
= − + s exp(−s t) |0 + s exp(−s t) dt
dt 2 dt 0 dt
t=0

d2 fi (t) dfi (t) ∞ dfi (t)
= − −s + s2 exp(−s t) dt
dt 2 dt t=0 0 dt
t=0

d2 fi (t) dfi (t)
= − 2
−s + s2 exp(−s t) fi (t) |∞ 3
0 + s s [fi (t)]
dt dt t=0
t=0

0 d2 fi (t) dfi (t)
= −s −s − s2 fi (t)t=0 + s3 s [fi (t)].
dt 2 dt t=0
t=0
n
d fi (t)
Next proceed by induction and write s n as follows :
dt
n
d fi (t) 0 dn−1 fi (t) dn−2 fi (t) 2 dn−3 fi (t)
s = −s −s −s
dt n dt 2 dt n−2 dt n−3
t=0 t=0 t=0
dfi (t)
− ... − sn−2 − sn−1 fi (t)t=0 + sn s [fi (t)] (5.88)
dt t=0
5.5 Computation 95
t
(8) : Consider f (t) = 0 fi (x)dx.
To solve (8) write:
t ∞ t
s fi (x)dx = exp(−s t) fi (x)dx dt ,
0 0 0
t=∞
exp(−s t) t 1 ∞
= fi (x)dx + exp(−s t)fi (t)dt
−s 0 t=0 s 0

s fi (t) Fi (s)
= 0+ = . (5.89)
s s
(9) : Consider f (t) = cos(a t). It is convenient to use the representation
exp(I a t) + exp(−I a t)
f (t) = cos(a t) = ,
2
and work with s > a.
∞
1 1 ∞
s [cos(a t)] = exp[(I a − s) t)] dt +
exp[(−I a − s) t)] dt ,
0 2 2 0

1 exp[(I a − s) t] ∞ 1 exp[(−I a − s) t] ∞
= +
2 Ia−s 0 2 −I a − s 0
1 −1 1 −1 s
= + = 2 . (5.90)
2 I a − s 2 −I a − s s + a2
(10) : Consider f (t) = sin(a t). It is convenient to use the representation
exp(a t) − exp(−a t)
sin(a t) =
2I
Work with s > a and write:
t t
1
s [sin(a t)] = exp[(a − s) t)] dt − exp[(−a − s) t)] dt ,
2I 0 0
∞ ∞
1 exp[(a − s) t] 1 exp[(−a − s) t]
= −
2I a−s 0 2I −a − s 0
1 −1 1 −1 a
= − = (−I ) 2 . (5.91)
2I a − s 2I −a − s s − a2
5.6 Heaviside Step Function
A slightly more detailed description of H (t − a) is given below. The Heaviside step

function (HSF) is defined as follows for a ≥ 0.
H (t − a) ≡ u(t − a) ≡ ua (t) = 1 for t ≥ a

= 0 for t < a . (5.92)
5.6.1 On- and Off Switches
The (HSF) is useful for representing switches that turn on and off at specific times.
For instance, consider the following two problems.
Using the notation of (5.1) in the form
V (t) Y (t) = F(t)
define various switches by choosing their inhomogeneous term F(t).

(HSF)-1 : First, if one is using a very simple switch that is on for t > 0 with a
value of W0 but turns off at t ≥ 5 then one has
F(t) = W0 , if t < 5
= 0 , if t ≥ 5 . (5.93)
In terms of the Heaviside function, this process is displayed as
F(t) = W0 {1 − H (t − 5)} . (5.94)
(HSF)-2 : On the other hand, a more sophisticated switch is one that is:
on for t < 5 with value W1 ;
goes off during 10 > t ≥ 5;
comes back on at t = 10 with value W2
and stays on until t = 15.
Turns off at t = 15
and comes back on at t = 20 with strength W3
and stays on until t = 25.
But at t = 25 the switch instantly adjusts to strength W4
and stays on at that strength.
Thus, one works with a switch that operates with function F(t) such that
5.6 Heaviside Step Function 97
F(t) = W1 , if t < 5
= 0 , if 5 ≤ t < 10
= W2 , if 10 ≤ t < 15
= 0 , if 15 ≤ t < 20
= W3 , if 20 ≤ t < 25
= W4 , if t ≥ 25 . (5.95)
In terms of the Heaviside function, this process can be displayed as follows:
F(t) = W1 {1 − H (t − 5)} + W2 {H (t − 10) − H (t − 15)}

+ W3 {H (t − 20) − H (t − 25)} + W4 {H (t − 25)} . (5.96)
5.7 Solving Initial Value Problems
The process:
One works out the Laplace transform of both sides of the given differential equation.
The initial conditions are inserted into the Laplace transformed equation. Generally,
this simplifies the output variable a little bit. Next, one reorganizes the output variable
by partial fraction decomposition. Then, one inverse Laplace transforms the resultant
output, if needs be, by using Laplace transform inversion tables. This process will
become clear as we work out several problems.
5.8 First-Order Differential Equations
Solution: (1)
Let us solve the following first-order differential equation with boundary condition
x(t) = 0 for t < 0.
dx(t)
(1) : 2 + 5 x(t) = t exp(−t) . (5.97)
dt
Solution of (1).
Laplace transform of both sides of differential equation (5.97)—namely equa-
tion (1)—are
1
2 {s X (s) − x(0)} + 5 X (s) = . (5.98)
(s + 1)2
A more organized form of (5.98) is
1 1 2 x(0)
X (s) = + . (5.99)
2s + 5 (s + 1)2 (2 s + 5)
To satisfy the initial boundary condition, set x(0) = 0.
1 1
X (s) = . (5.100)
2s +5 (s + 1)2
Notice how Laplace transform converts a function of some variable—say, t—into a

function of another variable—say, s.
5.8.1 Partial Fraction Decomposition of (5.100)
To decompose (5.100) into partial fractions express X (s) as follows:
A B C
X (s) ≡ + + (5.101)
(2 s + 5) (s + 1) (s + 1)2
Multiply the right-hand sides of (5.100) and (5.101)

by (2 s + 5) (s + 1)2 and equate the two results. [Note the objective here is to get
unity on the left-hand side of the following (5.102)]. We get
1 = A (s + 1)2 + B (s + 1) (2 s + 5) + C (2 s + 5)
= s2 (A + 2 B) + s (2 A + 2 C + 7 B) + (A + 5 C + 5 B). (5.102)
For (5.102) to hold for arbitrary values of s, terms with any particular power of s
must be equal on both sides of this equation. Accordingly, comparison of the s2 , s,
and s0 terms gives
A+2B = 0 ;
2A + 2C + 7B = 0 ;
A + 5C + 5B = 1 . (5.103)
Equation (5.103) is readily solved and the result is
4 2 1
A = ; B = − ; C = . (5.104)
9 9 3
5.8 First-Order Differential Equations 99
Accordingly, using (5.101) and (5.104), X (s) is rewritten as
4 1 2 1 1 1
X (s) = − +
9 (2 s + 5) 9 (s + 1) 3 (s + 1)2
4 1 2 1 1 1
= − + . (5.105)
18 (s + 5/2) 9 (s + 1) 3 (s + 1)2
Finally, by using the inverse Laplace transform tables, one inverts the above from
X (s) to x(t). The result is
4 5t 2 t
x(t) = exp − − exp(− t) + exp(− t) . (5.106)
18 2 9 3
Checking Accuracy of the Result (5.106)

In order to check x(t), as per (5.106), one uses the homogeneous part of the parent
differential equation (5.97), namely 2 dx(t)
dt
+ 5 x(t) and checks to see whether it
equals the inhomogeneous part, i.e., t exp(−t).
First, we calculate dx(t)
dt
.
dx(t) 10 5t 5 t
=− exp − + exp(− t) − exp(− t) , (5.107)
dt 18 2 9 3
Next, we calculate 2 dx(t)

dt
+ 5 x(t) and find it is equal to t exp(− t), as it must if
the solution is correct.
dx(t) 10 5t 10 2t
2 + 5 x(t) = − exp − + exp(− t) − exp(− t)
dt 9 2 9 3
20 5t 10 5t
+ exp − − exp(− t) + exp(− t)
18 2 9 3
= t exp(− t) . Q E D . (5.108)
Solution: (2)
Again, with the same initial boundary condition, namely x(t) = 0 for t < 0, let us
solve another first-order differential equation: namely (5.109).
dx(t)
(2) : 3 − 7 x(t) = cosh(a t). (5.109)
dt
Solution of (2).
Laplace transform of both sides of the differential equation (5.109)—namely
equation (2)— are
s
3 {s X (s) − x(0)} − 7 X (s) = . (5.110)
(s2 − a2 )
Or
1 s 3 x(0)
X (s) = + . (5.111)
(s2 − a2 ) 3s − 7 (3 s − 7)
To satisfy the initial boundary condition, set x(0) = 0.
1 1 s
X (s) = . (5.112)
(s − a) (s + a) 3s − 7

To decompose (5.112) into partial fractions, express X (s) as follows:
A B C
X (s) ≡ + + (5.113)
(s − a) (s + a) (3s − 7)

by (s − a) (s + a) (3 s − 7) and equate the two results. [Note the objective here is to
get s on the left-hand side.] We get
s = A (s + a) (3 s − 7) + B (s − a) (3 s − 7) + C (s − a)(s + a)
= s2 (3 A + 3 B + C) + s (3 a A − 7 A − 7 B − 3 a B)
+ (−7 a A + 7 a B − a2 C) . (5.114)
If (5.114) is to hold for arbitrary values of s, terms with any particular power of s
and s0 terms gives
3A + 3B + C = 0 ;
(3 a A − 7 A − 7 B − 3 a B) = 1;
(−7 a A + 7 a B − a2 C) = 0. (5.115)
3a + 7 3a − 7 42
A = ; B = − ; C = − (5.116)
18 a2 − 98 18 a2 − 98 18 a2 − 98
Accordingly, using (5.113) and (5.116), one calculates X (s).
1 3a + 7 1 3a − 7
X (s) = −
(s − a) 18 a2 − 98 (s + a) 18 a2 − 98
1 42
− . (5.117)
(3 s − 7) 18 a2 − 98
X (s) to x(t). The answer is
3a + 7 3a − 7
x(t) = exp(a t) − exp(− a t)
18 a − 98
2 18 a2 − 98
7t 14
− exp . (5.118)
3 18 a2 − 98
Checking Accuracy of the Result x(t)

To that purpose, for x(t) as per (5.118), one uses the homogeneous part of the parent
differential equation (5.109), namely 3 dx(t)
dt
− 7 x(t), and checks to see whether it
equals the inhomogeneous part cosh(a t).
We get
dx(t)
3 − 7 x(t)
dt
3a + 7 3a − 7
= 3a exp(a t) + 3 a exp(− a t)
18 a − 98
2 18 a2 − 98
98 7 3a + 7
− exp −7 exp(a t)
18 a − 98
2 3t 18 a2 − 98
3a − 7 98 7
+7 exp(− a t) + exp . (5.119)
18 a − 98
2 18 a − 98
2 3t
7
Combining exp(a t), exp(− a t), and the exp 3t
terms, (5.120) leads to the
expression
dx(t)
3 − 7 x(t)
dt
(3 a + 7) (3 a − 7) (3 a − 7) (3 a + 7)
= exp(a t) + exp(−a t)
18 a2 − 98 18 a2 − 98
98 7 98 7
− exp + exp (5.120)
18 a − 98
2 3t 18 a − 98
2 3t
with the result


dx(t) exp(a t) + exp(− a t)
3 − 7 x(t) = = cosh(a t). (5.121)
dt 2
Q.E.D.
Solution: (3)
Another first-order differential equation with boundary condition x(t) = 0 for t < 0
is solved below.
dx(t)
(3) : 5 + 4 x(t) = sinh(a t) . (5.122)
dt
Solution of (3).
tion (3)—are
a
5 {s X (s) − x(0)} + 4 X (s) = . (5.123)
s2 − a 2
Or
a
X (s) (5 s + 4) − 5 x(0) = (5.124)
s2 − a2
Note: We have used the Laplace transform table provided in (5.78). To satisfy the
initial boundary condition, set x(0) = 0. As a result, X (s) becomes

a
X (s) = . (5.125)
(5 s + 4) (s − a) (s + a)

A B C
X (s) ≡ + + (5.126)
(5 s + 4) (s − a) (s + a)

by (5 s + 4) (s − a) (s + a) and equate the two results. [Note the objective here is to
get a on the left-hand side of (5.127). We get
a = A (s − a) (s + a) + B (5 s + 4) (s + a) + C (5 s + 4)(s − a) . (5.127)
and s0 terms gives
A + 5B + 5C = 0 ;
5aB + 4B − 5aC + 4C = 0 ;
−A a2 + 4 a B − 4 a C = a . (5.128)

5 1 1 1 1
A = − ; B = ; C = − .
2 4 − 5a 4 + 5a 2 (4 + 5 a) 2 (4 − 5 a)
(5.129)
By using (5.126) and (5.129), X (s) is rewritten as

20 1 1
X (s) = + − .
(5 s + 4) (16 − 25 a ) (8 + 10 a) (s − a) (8 − 10 a) (s + a)
2
(5.130)
X (s) to x(t). The result is
4 4t 1 1
x(t) = exp − + exp(a t) − exp(− a t) .
(16 − 25 a2 ) 5 (8 + 10 a) (8 − 10 a)
(5.131)

In order to check the accuracy of the result x(t), given as per (5.131),
one uses the homogeneous part of the parent differential equation (5.122), namely
5 dx(t)
dt
+ 4 x(t), and checks to see whether it equals the inhomogeneous part
sinh(a t).
To that end, first one calculates 5 dx(t)
dt
.
dx(t) 16 4t 5a
5 =− exp − + exp(a t)
dt 16 − 25 a2 5 8 + 10 a
5a
+ exp(− a t) . (5.132)
8 − 10 a
Next one calculates 4 x(t).

16 4t 2
4 x(t) = exp − + exp(a t)
16 − 25 a2 5 4 + 5a
2
− exp(− a t) . (5.133)
4 − 5a
Upon adding, (5.132) and (5.133) one can write
dx(t)
5 + 4 x(t)
dt
5a 4
= exp(a t) + exp(a t)
8 + 10 a 8 + 10 a
5a 4
+ exp(− a t) − exp(− a t) (5.134)
8 − 10 a 8 − 10 a
Equation (5.134) leads readily to the desired result.
dx(t)
5 + 4 x(t)
dt
1
= {exp(a t) − exp(− a t)} = sinh(a t). (5.135)
2
Q.E.D.
Solution: (4)
Another first-order differential equation with boundary condition x(t) = 0 for t < 0
is solved below.
dx(t)
(4) : 4 + 9 x(t) = exp(a t) sinh(a t) . (5.136)
dt
Solution of (4).
tion (4)—are
a
4 {s X (s) − x(0)} + 9 X (s) = . (5.137)
s (s − 2 a)
Or
a
X (s) (4 s + 9) − 4 x(0) = (5.138)
s (s − 2 a)
Note: We have used the Laplace transform table provided in (5.78). To satisfy the
initial boundary condition, set x(0) = 0. As a result, X (s) becomes

a
X (s) = . (5.139)
(s) (s − 2 a) (4 s + 9)

A B C
X (s) ≡ + + (5.140)
(s) (s − 2 a) (4 s + 9)
by (s) (s − 2 a) (4 s + 9) and equate the two results. [Note the objective here is to
get a on the left-hand side of (5.141). We get
a = A (s − 2 a) (4 s + 9) + B (s) (4 s + 9) + C (s) (s − 2 a)

= A 4 s2 + 9 s − 8 a s − 18 a + B 4 s2 + 9 s + C s2 − 2 a s . (5.141)
and s0 terms gives
4A + 4B + C = 0 ;
9A − 8aA + 9B − 2aC = 0 ;
−18 a A = a . (5.142)

1 1 2 4
A = − ; B = ; C = − + .
18 18 + 16 a 9 + 8a 18
(5.143)
By using (5.140) and (5.143), X (s) is rewritten as

−1 1 1 4 2 1
X (s) = + + − .
18 s 18 + 16 a s −2a 18 9 + 8a 4s + 9
(5.144)
Finally, by using the inverse Laplace transform tables, one inverts X (s) to x(t). The
result is

1 exp(2 a t)
x(t) = − +
18 18 + 16 a

4 2 1 9t
+ − exp − . (5.145)
18 9 + 8 a 4 4

In order to check the accuracy of the result x(t), given as per (5.145), one uses the
homogeneous part of the parent differential equation (5.136), namely
4 dx(t)
dt
+ 9 x(t), and checks to see whether it equals the inhomogeneous part exp(a t)
sinh(a t). To that end, first one calculates 4 dx(t)
dt
.

dx(t) 4 a exp(2 a t) 9 2 2 9t
4 = − − exp − . (5.146)
dt 9 + 8a 4 9 9 + 8a 4
And next 9 x(t).

1 9 exp(2 a t) 9 2 2 9t
9 x(t) = − + + − exp − .
2 2 (9 + 8 a) 4 9 9 + 8a 4
(5.147)
Upon adding (5.146) and (5.147), one gets

dx(t) 4 a exp(2 a t) 1 9 exp(2 a t)
4 + 9 x(t) = − + . (5.148)
dt 9 + 8a 2 2 (9 + 8 a)
Equation (5.148) leads readily to the desired result.
dx(t) exp (2 a t) − 1
4 + 9 x(t) = . (5.149)
dt 2
Q.E.D.
Solution: (5)
Another first-order differential equation with boundary condition x(t) = − 6 for
t < 0 is solved below.
dx(t)
(5) : + 5 x(t) = exp(7 t) . (5.150)
dt
Solution of (5).
tion (5)—are
1
{s X (s) − x(0)} + 5 X (s) = . (5.151)
(s − 7)
Note: We have used the Laplace transform table provided in (5.78) and to satisfy the
initial boundary condition, we have set x(0) = − 6. As a result, X (s) becomes
1
X (s) (s + 5) + 6 = . (5.152)
(s − 7)
Equivalently,

1 6 43 − 6 s
X (s) = − = . (5.153)
(s + 5) (s − 7) s+5 (s + 5) (s − 7)
A B
X (s) ≡ + . (5.154)
(s + 5) (s − 7)

by (s + 5) (s − 7) and equate the two results. We get
43 − 6 s = A (s − 7) + B (s + 5) = s (A + B) + (5 B − 7 A) . (5.155)
must be equal on both sides of this equation. Accordingly, comparison of the s and
s0 terms gives
A+B = − 6 ;
−7 A + 5 B = 43 (5.156)
leading to
73 1
A = − ; B = . (5.157)
12 12
With the help of (5.154) and (5.157), X (s) is rewritten as
−73 1 1
X (s) = + . (5.158)
12 (s + 5) 12 s−7
Finally, by using the inverse Laplace transform tables, one inverts X (s) to x(t). The
result is
73 1
x(t) = − exp(−5 t) + exp(7 t) . (5.159)
12 12

The result x(t), as given in (5.159), is accurate because, it satisfies the primary
differential equation (5.150). To check proceed as follows:
dx(t)
+ 5 x(t)
dt
73 7
=5 exp(−5 t) + exp(7 t)
12 12
73 5
−5 exp(−5 t) + exp(7 t) = exp(7 t) . (5.160)
12 12
Q.E.D.
5.9 Second-Order Differential Equations
5.9.1 Solution by Laplace transform
By the use of Laplace transform, solve the following second-order differential equa-
tion
(I ) : 2 y (t) + 3 y (t) − 2 y(t) = t exp(−2 t) (5.161)
with the boundary condition
y(0) = 0 and y (0) = −2 . (5.162)
Solution: (I)
tion (I)—are
1
2 s2 Y (s) − s y(0) − y (0) + 3 {s Y (s) − y(0)} − 2 Y (s) = .
(s + 2)2
(5.163)
Upon inserting the boundary condition (5.162), (5.163) becomes
1
Y (s) 2 s2 + 3 s − 2 = Y (s) [(2 s − 1) (s + 2)] = − 4. (5.164)
(s + 2)2
Equation (5.164) can be rewritten in a more compact form as
− 4 s2 − 16 s − 15
Y (s) = . (5.165)
(2 s − 1) (s + 2)3
5.9 Second-Order Differential Equations 109
To decompose (5.165) into partial fractions, express Y (s) as follows:
A B C D
Y (s) ≡ + + + (5.166)
(2 s − 1) (s + 2) (s + 2)2 (s + 2)3

by (2 s − 1) (s + 2)3 and equate the two results. We get
− 4 s2 − 16 s − 15
= A (s + 2)3 + B (2 s − 1) (s + 2)2 + C (2 s − 1) (s + 2) + D (2 s − 1)
= (A + 2 B) s3 + (6 A + 7 B + 2 C) s2 + (12 A + 4 B + 3 C + 2 D) s
+ (8A − 4B − 2C − D) . (5.167)
For (5.167) to hold for arbitrary values of s, terms with any particular power of s must
be equal on both sides of this equation. Accordingly, comparison of the s3 , s2 , s,
and s0 terms gives
A +2B = 0 ;
6A + 7B + 2C = − 4 ;
12 A + 3 C + 4 B − 2 D = − 16 ;
8 A − 4 B − 2 C − D = − 15 . (5.168)
Equation (5.168) is straightforward to solve and the result is
192 96 10 25
A = − ; B = ; C = − ; D = − . (5.169)
125 125 125 125
There is a common denominator of 125 for A, B, C, and D in (5.169). Therefore
upon plugging these results for A, B, C, and D into the Laplace transform (5.166)
one gets

1 96 96 10 25
Y (s) = − + − − . (5.170)
125 (s − 2 ) (s + 2) (s + 2)
1 2 (s + 2)3
Y (s) to y(t). The result is

1 t 25 2
x(t) = −96 exp + 96 exp(−2 t) − 10 t exp(−2 t) − t exp(−2 t)
125 2 2
(5.171)

In order to confirm the accuracy of x(t), as per (5.171), one uses first the homogeneous
part of the parent differential equation (5.161), namely 2 y (t) + 3 y (t) − 2 y(t), and
checks to see whether after using the boundary condition y(0) = 0 and y (0) =
−2 it equals the inhomogeneous part of the parent differential equation, i.e.,
t exp(−2 t). To that purpose, let us start in seriatim with the first two terms of
x(t) recorded in (5.171). That is,
2
96 d d
2 2 + 3 − 2 . {− exp(t/2) + exp(−2 t)}
125 dt dt

96 2 3
= − − + 2 exp(t/2) + (4 − 6 − 2) exp(−2 t) = 0 . (5.172)
125 4 2
Next, we examine the response of the third and the fourth terms in (5.171). Meaning
the expressions in the following two equations:
2
10 d d
− 2 2 + 3 − 2 [t exp(−2 t)] , (5.173)
125 dt dt
2
25 d d
− 2 2 + 3 − 2 [t 2 exp(−2 t)] . (5.174)
2 × 125 dt dt
With a little bit of simple differentiation and great deal of straightforward

algebra,
(5.173) and (5.174) yield the results 25 exp(−2 t) and exp(−2 t) t − 25 . Adding the
two gives t exp(−2 t). Q.E.D.
(II)
With the help of Laplace transform, solve the following second-order differential
equation
(II ) : y (t) − 6 y (t) + 9 y(t) = 2 sin(3 t). (5.175)
The boundary condition is:
y(0) = −1 and y (0) = −4 . (5.176)
Solution: (II)
tion (II)—are
6
s2 Y (s) − s y(0) − y (0) − 6 {s Y (s) − y(0)} + 9 Y (s) = .
(s2 + 9)
(5.177)
6
Y (s) s2 + 6 s + 9 + s − 2 = (5.178)
s2 + 9
After a little bit of algebra equation (5.178) can be rewritten in a more compact form.
−s3 + 2 s2 − 9 s + 24
Y (s) = . (5.179)
(s2 + 9) (s − 3)2
As + B C D
Y (s) ≡ + + . (5.180)
(s + 9) (s − 3) (s − 3)2
2

by (s2 + 9) (s − 3)2 and equate the two results. We get
− s3 + 2 s2 − 9 s + 24 = (A s + B)(s − 3)2 + C (s2 + 9) (s − 3) + D (s2 + 9) .

(5.181)
For (5.181) to hold for arbitrary values of s, terms with any particular power
of s must be equal on both sides of this equation. Accordingly, comparison of the
s3 , s2 , s, and s0 terms gives
A+C = − 1 ;
−6A + B − 3C + D = 2 ;
−6B + 9A + 9C = − 9 ;
9 B − 27 C + 9 D = 24 . (5.182)
1 10 6
A = ; B = 0 ; C = − ; D = − . (5.183)
9 9 9
While B = 0, there is a common denominator of 9 for A, C, and D in (5.183). There-
fore upon plugging these results for A, B, C, and D into the Laplace transform (5.180)
one gets

1 s 10 6
Y (s) = − − . (5.184)
9 (s2 + 9) (s − 3) (s − 3)2
Y (s) into Y (t). The result is
1
Y (t) = cos(3 t) − 10 exp(3 t) − 6 t exp(3 t) . (5.185)
9

In order to confirm the accuracy of Y (t), as per (5.185), we use the parent differential
equation (5.175), namely Y (t) − 6 Y (t) + 9 Y (t) , and check to see whether after
using the boundary condition y(0) = − 1 and y (0) = −4 it equals the
inhomogeneous part of the parent differential equation, namely 2 sin(3 t). It does
and the relevant algebra is described below.
Y (t) − 6 Y (t) + 9 Y (t)

= − cos(3 t) − 14 exp(3 t) − 6 t exp(3 t)

3 10 2
− 6 − sin(3 t) − exp(3 t) − exp(3 t) − 2 t exp(3 t)
9 3 3
+ cos(3 t) − 10 exp(3 t) − 6 t exp(3 t) .
= 2 sin(3 t) . (5.186)
Q.E.D.
With the help of Laplace transform, solve the following IVP.
(III ) : y (t) − 8 y (t) + 7 y(t) = 9 t. (5.187)
The boundary condition is
y(0) = −1 , y (0) = 2 . (5.188)
Solution: (III)
tion (III)—are
9
s2 Y (s) − s y(0) − y (0) − 8 {s Y (s) − y(0)} + 7 Y (s) = 2 . (5.189)
s
9
Y (s) s2 − 8 s + 7 + s − 10 = 2 (5.190)
s
Equation (5.190) can readily be rewritten in a more compact form.
−s3 + 10 s2 + 9
Y (s) = . (5.191)
s2 (s − 7) (s − 1)
To decompose (5.191) into partial fractions, we express Y (s) as follows:
A B C D
Y (s) ≡ + 2+ + . (5.192)
s s (s − 7) (s − 1)

by s2 (s − 7) (s − 1) and equate the two results. We get
− s3 + 10 s2 + 9 = A s (s − 7) (s − 1) + B (s − 7) (s − 1)
+ C s2 (s − 1) + D s2 (s − 7) . (5.193)
For (5.193) to hold for arbitrary values of s, terms with any particular power
of s must be equal on both sides of this equation. Accordingly, comparison of the
s3 , s2 , s, and s0 terms gives
A+C +D = −1 ;
− 8 A + B − C = 10 ;
7A − 8B − 7D = 0 ;
7B = 9 . (5.194)
72 9 26
A = ; B = ; C = ; D = − 3. (5.195)
49 7 49
Upon plugging the results for A, B, C, and D into the Laplace transform (5.192) one
gets
72 9 26
3
Y (s) = 49
+ 7
+ 49
− . (5.196)
s s2 (s − 7) (s − 1)
Y (s) to Y (t). The result is
72 9 26
Y (t) = + t+ exp(7 t) − 3 exp(t). (5.197)
49 7 49

In order to confirm the accuracy of Y (t), as per (5.197), we use the parent differential
equation (5.187), namely Y (t) − 8 Y (t) + 7 Y (t) , and check to see whether after
using the boundary condition y(0) = − 1 and y (0) = 2, it equals the inhomoge-
neous part of the parent differential equation, namely 9 t. It does, and the relevant
algebra is displayed neatly below.
Y (t) − 8 Y (t) + 7 Y (t)

= 26 exp(7 t) − 3 exp(t)

72 8 × 26
+ − − exp(7 t) + 24 exp(t)
7 7

72 26
+ + 9t + exp(7 t) − 21 exp(t) .
7 7
= 9t . (5.198)
Q.E.D.
With the help of Laplace transform, solve the following differential equation.
(I V ) : 2 y (t) − 5 y (t) − 3 y(t) = t exp(− t). (5.199)
The boundary condition is
y(0) = 1 , y (0) = 2 . (5.200)
Solution: (IV)
tion (IV)—are
1
2 s2 Y (s) − s y(0) − y (0) − 5 {s Y (s) − y(0)} − 3 Y (s) = .
(s + 1)2
(5.201)
1
Y (s) 2 s2 − 5 s − 3 − 2 s − 4 + 5 = (5.202)
(s + 1)2
Equation (5.202) can readily be rewritten in a more compact form.
(2 s − 1) (s + 1)2 + 1
Y (s) = . (5.203)
(s + 1)2 (2 s + 1) (s − 3)
A B C D
Y (s) ≡ + + + . (5.204)
(2 s + 1) (s − 3) (s + 1) (s + 1)2

by (s + 1)2 (2 s + 1) (s − 3) and equate the two results. As usual do the relevant
algebra and determine the parameters A, B , C, etc. [Note : For instance, compare
with (5.192) to (5.195).] We get
236 1 9 1
A= ; B= ; C= ; D= . (5.205)
7 × 16 7 × 16 16 4
Upon plugging the above results for A, B, C, and D into the Laplace trans-
form (5.204) one gets
59 1 1 1
Y (s) = +
28 (2 s + 1) 7 × 16 (s − 3)
9 1 1 1
+ + . (5.206)
16 (s + 1) 4 (s + 1)2
Y (s) to Y (t). The result is
59 1 1
Y (t) = exp(− t/2) + exp(3 t)
28 2 7 × 16
9 t
+ exp(− t) + exp(− t) . (5.207)
16 4

In order to determine the accuracy of Y (t), as per (5.207), we use the parent differen-
tial equation (5.201), namely 2 Y (t) − 5 Y (t) − 3 Y (t), and check to see whether
after using the boundary condition y(0) = 1 and y (0) = 2, the result equals the
inhomogeneous part of the parent differential equation, namely t exp(− t). We find
that it does, and the relevant algebra is displayed in seriatim below in (5.208), (5.209),
and (5.210).
1 59 9 9
2 Y (t) = exp(− t/2) + exp(3 t) + exp(− t)
4 28 56 8
1
− exp(− t) + t exp(− t) . (5.208)
2
5 59 15 45
− 5 Y (t) = exp(− t/2) − exp(3 t) + exp(− t)
4 28 7 × 16 16
5 5
− exp(− t) + t exp(− t) . (5.209)
4 4
59 3 27
− 3 Y (t) = −3 exp(− t/2) − exp(3 t) − exp(− t)
56 7 × 16 16
3
− t exp(− t) . (5.210)
4
Q.E.D.
Adding (5.208), (5.209), and (5.210) leads to the desired
result t exp(− t). To double check this statement, examine the details of the addition
as given below. We have

59 1 5
exp(− t/2) + −3 = 0.
56 2 2

9 15 3
exp(3 t) − − = 0.
56 7 × 16 7 × 16
1
exp(− t) [18 − 16 + 45 − 20 − 27] = 0 .
16
1
t exp(− t) [2 + 5 − 3] = t exp(− t) . (5.211)
4
‘Quod Erat Demonstrandum.’
5.10 Need for Convolution
We have studied inhomogeneous linear ordinary differential equations of the form
O(t) P(t) = (t) (5.212)
where the differential operator O(t), the solution P(t), as well as the inhomoge-
neous term represented by a function (t), all involved constants and derivatives
with respect to a single variable t. In addition, there were some specified boundary
conditions that the solution had to satisfy. The important thing to note was that the
inhomogeneous term L(t) was not arbitrary. Rather, it was properly defined.
In preceding chapters, various methods for solving homogeneous linear ordinary
differential equations were discussed.
5.10 Need for Convolution 117
For inhomogeneous linear ordinary differential equations, those methods required

working out the particular integral, Ipi (t), ab initio for every different inhomogeneous
term (t). In this regard, Green’s function methodology appeared as a savior, for the
reason that Green’s function—if successfully calculated—once and for all, helps
solve the differential equation for arbitrary values of the inhomogeneous term (t), .
Consequently, for any particular differential operator O(t), in addition to the homo-
geneous solution, only Green’s function itself needs to be worked out.
Unfortunately, an appropriate Green’s function sometimes is hard to work out.
And even when it is not, the solution is likely to contain infinite terms and be very
inconvenient to use.
For instance, the eigenfunction Green’s function leads to infinite series that are
often both difficult to work out and complicated to work with. Even the closed-form
solutions of Green’s function, that make use of Dirac’s delta function, are complicated
to calculate and can be employed only to a limited class of Green’s functions.
With the hope of remedying these difficulties, one uses convolution integrals
which are useful for solving differential equations with arbitrary values of the inho-
mogeneous term—namely the forcing function—(t).
5.11 Convolution Integral
Given continuous, or at least piecewise continuous, functions f (t) and g(t) on [0, ∞] ,
the convolution integral of f (t) and g(t) is defined as
t t
f (t) (∗)g(t) = f (t − τ ) g(τ ) dτ = g(t − τ ) f (τ ) dτ . (5.213)
0 0
It is helpful to observe how a convolution integral is employed for solving differen-

tial equations with an arbitrary inhomogeneous term . To that end, let us solve the
following very simple IVP.
y (t) + y(t) = γ(t) . (5.214)
Its Laplace transform is

Y (s) s2 − s y(0) − y (0) + Y (s) = γ(s) (5.215)
Using the boundary condition
y(t = 0) = 1 , y (t = 0) = − 6 , (5.216)
(5.215) leads to
s 6 γ(s)
Y (s) = − + . (5.217)
(s2 + 1) (s2 + 1) (s2 + 1)
Finally, taking the inverse transform of (5.217) one gets

t
y(t) = cos(t) − 6 sin(t) + sin(τ ) γ(t − τ ) dτ . (5.218)
0
The main idea of this exercise is that the above result applies to arbitrary choices
of γ(t). Just to make sure that we have not made an error somewhere let us make
an extremely simple choice for the arbitrary γ(t). That is, γ(t) = t. Then, (5.218)
gives
t
y(t) = cos(t) − 6 sin(t) + sin(τ ). (t − τ ) dτ . (5.219)
0
Now integrate the following by parts

t
sin(τ ). (t − τ ) dτ . (5.220)
0
One gets
t t
sin(τ ). (t − τ ) dτ = [− cos(τ ) (t − τ )]t0 − cos(τ ) d τ
0 0
− cos(t) [t − t] + cos(0) [t − 0] − [sin(t) − sin(0)]
= t − sin(t) . (5.221)
Equations (5.219) to (5.222) lead to the final result.
y(t) = cos(t) − 6 sin(t) + [t − sin(t)] .

= cos(t) − 7 sin(t) + t. (5.222)
One can quickly confirm that this result is correct because
y (t) + y(t) = − cos(t) + 7 sin(t) + cos(t) − 7 sin(t) + t

= γ(t) = t. (5.223)
Chapter 6
Special Types of Differential Equations
One special type of differential equation, namely the Ber noulli 1. equation, was
discussed in Chap. 4. [Compare, for instance, (4.19)–(4.77).] Here, that analysis is
extended to other special type equations.
Included in this presentation are the Clairaut 2. equations—[Compare (6.2)–
(6.13)]— Lagrange3. equation—[Compare (6.19)–(6.31)], y the separable equations—
[Compare (6.32)–(6.35)], and the dy dx
= x
equations—[Compare
(6.36)–(6.73)]. In addition, there are the so-called exact [Compare (6.74)–(6.91)] and
inexact equations—[Compare (6.92)–(6.241)]—Riccati 4. equations—[Compare
(6.242)–(6.268)]—Euler 7. equations—[Compare (6.269)–(6.315)], and the fac-
torable equations—[Compare (6.316)–(6.344)].
Notation
Occasionally, for convenience, the following notation will be used.
dy
q= ;
dx
q dx = dy . (6.1)
6.1 Clairaut Equation: Description
Consider an equation
y = σ x + f (σ) (6.2)
where y is the dependent, x is the independent variable, and f (σ) is a function of an

arbitrary parameter σ. Differentiate (6.2) with respect to x,

https://doi.org/10.1007/978-3-319-76406-1_6
120 6 Special Types of Differential Equations
dy
= σ,
dx
and use (6.1) in the form
dy
= q,
dx
that leads to
σ = q .
Now substitute the variable q for the parameter σ in (6.2). This process leads to
Clairaut equation
y = q x + f (q) . (6.3)
Its general solution has already been recorded as (6.2). It is a straight line in the (x, y)
plane that is obtained by replacing the slope, q, by its observed value σ. Indeed, the
general solution is a single arbitrary parameter representation of a whole family of
straight lines.
6.1.1 Solving Clairaut Equation
Consider the Clairaut equation (6.3). Differentiate both sides.
dy = q dx + x dq + d f (q) .
According to (6.1), q dx = dy. Thus, dy and q dx can be eliminated from the above
differential equation. The result is
x dq + d f (q) = 0 . (6.4)
It is convenient to introduce the notation
d f (q) = F(q) dq , (6.5)
whereby (6.4) can be written as
[x + F(q)] · dq = 0 . (6.6)
Notice that (6.6) is a product of two differential equations:

6.1 Clairaut Equation: Description 121
(1) : dq = 0 ;
(2) : [x + F(q)] = 0 . (6.7)
The differential equation (6.7)-(1) is straightforward and can immediately be inte-

grated. Initial integration gives

dq = q = 0 = σ ,
where σ is an arbitrary constant. The next integration yields

dy
q dx = dx = dy = y
dx

= σ dx = σ x + constant ,
which can be expressed as
y = σ x + f (σ) , (6.8)
where f (σ) is an arbitrary function of the arbitrary constant σ. Written in this form,
y is the general solution and is identical to the solution expressed by (6.2).
Clearly, the behavior of the second differential equation, namely (6.7)-(2), would
depend on the details of the function F(q). Relevant issues regarding this matter will
be analyzed in detail later.
Find general solution of the following Clairaut equation.

dy a
(A A) : y = x + dy
dx dx

a
= xq+ . (6.9)
q
According to the procedure explained above, general solution of a Clairaut equa-

tion is readily obtained by replacing the variable q—that is, dy
dx
—with parameter σ.
Thus, general solution of (A A) is
a
y =xσ+ . (6.10)
σ
Comment
Equation (6.10) describes a series of straight lines that could be assumed to be parallel
to tangents to the parabola
y2 = 4 a x . (6.11)
If this assumption turns out to be correct, these straight lines and tangents to the
parabola will have the same slope—meaning, at any given point (x, y) the value
of the differential dy
dx
derived from (6.11) will equal that obtained from the general
solution (6.10).
In order to check the above assumption, proceed√ as follows. Square root of the
equation of the parabola, (6.11), gives [y] parabola = 4 a x. Therefore
√
dy d 4ax a
= = . (6.12)
dx parabola dx x

dy
Now, replace dx
in Clairaut equation (A A) by dy
dx
—or equivalently replace
parabola
the existing σ in the solution (6.10) by dy
dx
. Either way, one gets
parabola
⎛ ⎞
a a
y=x + ⎝ ⎠
x a
x
√ √
=2 a x . (6.13)
And this is (the square root of) the equation of the parabola itself !
6.1.3 Singular Solution
Equation (6.13) is clearly a singular solution of the differential equation in example

(A A) because while it legitimately solves the differential equation, it is not obtained
by tweaking its general solution.
The success of the above exercise is due to fortunate geometrical intuition. Clearly,
successful intuition cannot be guaranteed. Therefore, it is important to devise an
algebraic procedure for determining singular solutions of Clairaut equations. Such
procedure is given below.
An Informal Procedure
(1) Begin with general solution y ≡ y(x) .
For instance, general solution for differential equation (A A) is (6.10).
(2) Differentiate general solution, y, with respect to σ. Set the result for dσ
dy
equal to
zero. Solve the resultant equation for σ, and notate the solution σ0 . For instance, for
dy
the given general solution (6.10), the differential is dσ = x − σa2 . Now, set it equal
to zero
dy a
=x− 2 = 0 (6.14)
dσ σ
a
and determine the solution for σ The result is σ = . Next, notate this σ σ0 . In
x
this manner, one has σ0 = ax .
(3) Replace σ by σ0 in the general solution.

For the present case that means

a
y = x σ0 +
σ0

a a
= x + a
x x
√
= 2 ax . (6.15)
Voila! The singular solution.

Squaring both sides of (6.15) leads to the parabola y 2 = 4ax.
Formal Procedure
For the Clairaut equation (6.3), the formally suggested second solution—noted in
(6.7)-(2)—is most likely the singular solution. It is helpful to determine whether this
likelihood holds true.
Comparing (6.3), (6.5), (6.7)-(2) and (6.9), one has f (q) = a
q
and
d f (q) = − qa2 dq. Accordingly, F(q) = − a
q2
and the suggested singu-
lar solution, [x + F(q)] = 0, that is

a
x − = 0,
q2
a
yields q = x
. Integration leads to

dy
q dx = dx = y
dx

a √
= dx = 2 a x .
x
And, as suspected, this is identical to the result in (6.15)—the solution obtained by

using the informal procedure! Thus either of the two, formal or informal, procedures
for determining the singular solution may be used.
6.1.4 Equations I(B)–I(F)
Find both general and singular solution to the following Clairaut equations numbered
I(B)–I(F).
I (B) : y = x q + 2q 2 .
I (C) : y = x q + 4q 3 .
I (D) : y = x q − 3 sin(q).
I (E) : y = x q + 5 cos(q).
I (F) : y = x q − sin2 (q). (6.16)
6.1.5 Informal Solution
I (B) : General Solution is : y = x σ + 2σ 2 .

dy x
Setting, = x + 4 σ = 0 , gives σ = σ0 = − .
dσ 4
2 2
−x −x x
Singular solution is : y = x +2 = − .
4 4 8
I (C) : General Solution is : y = x σ + 4σ 3 .

dy x
Setting = x + 12 σ 2 = 0 gives σ0 = − .
dσ 12
2
x x
Accor dingly, singular solution is : y = − x +4 − .
12 12
x 2x
Equivalently , y = − ; Or , 27y 2 + x 3 = 0 .
12 3
I (D) : General Solution is : y = x σ − 3 sin(σ)

dy x
Setting = x − 3 cos(σ) = 0 , gives σ0 = cos−1 .
dσ 3
x x
Singular solution is : y = x cos−1 − 3 sin cos−1 .
3 3
I (E) : General Solution is : y = x σ + 5 cos(σ).

dy x
Setting = x − 5 sin(σ) = 0 , gives σ0 = sin−1 .
dσ 5
x
Accor dingly, singular solution is : y − x sin−1 = 5 cos(σ0 ) .
5
I (F) : General Solution is : y = x σ − sin2 (σ) .

dy 1
Setting = x − sin(2 σ) = 0 , gives σ0 = sin−1 (x) .
dσ 2
x 1 −1
Singular solution is : y = sin−1 (x) − sin2 sin (x) . (6.17)
2 2
6.1.6 Formal Solution
d(2 q 2 )
F(q) = ;
dq
Equation I (B) : F(q) + x = 4 q + x = 0 .

dy x2
T her e f or e 4 dx = 4y = − (x) dx = − + const
dx 2
2
x
Singular solution is : y = − + constant .
8
d(4 q 3 )
F(q) = ;
dq
Equation I (C) : F(q) + x = 12 q 2 + x = 0 .

−1 √
T her e f or e qdx = dy = x dx.
12

−1 2 3
Singular solution is : y = x 2 + const.
12 3
d[−3 sin(q)]
F(q) = ;
dq
Equation I (D) : F(q) + x = − 3 cos(q) + x = 0 .

x
T her e f or e qdx = dy = cos−1 dx.
3
x
Singular solution is : y = x cos−1 − 9 − x 2 + const .
3
x x
Equivalently , y = x cos−1 − 3 sin cos−1 + const.
3 3
d[5 cos(q)]
F(q) = ;
dq
Equation I (E) : F(q) + x = − 5 sin(q) + x = 0 .

x
T her e f or e dy = sin−1 dx .
5
x
Singular solution is : y = x sin−1 + 25 − x 2 + const.
5
−1 x x
Equivalently , y = x sin + 5 cos sin−1 + const.
5 5
d[− sin2 (q)]

F(q) = ;
dq
Equation I (F) : F(q) + x = − sin(2 q) + x = 0

1
T her e f or e dy = sin−1 (x) dx
2

x 1
Singular solution is : y = sin−1 (x) + 1 − x 2 + const
2 2

x 1
Equivalently , y = sin−1 (x) − sin2 sin−1 (x) + const.
2 2
(6.18)
Solve the following Clairaut differential equations problems labeled (1)–(4). Find
both the general solution and the singular solution. [Hint: Read (6.10)–(6.18)]
y = x q + 3 q 2 . (1)
y = x q + 3 q 3 . (2)
y = x q − 2 sin q . (3)
y = x q + 2 cos q . (4)
6.2 Lagrange Equation
Similar to the Clairaut differential equation, but somewhat more general, is a

Lagrange equation that can be written as
6.2 Lagrange Equation 127
y = x G(q) + F(q) . (6.19)
If the function G(q) should be equal to q, Lagrange equation (6.19) would reduce
to Clairaut equation.
Differentiate (6.19)
dy = x dG(q) + G(q) dx + dF(q)
and use the following relationship proposed in (6.1)
dy = q dx .
The result is
q dx = x dG(q) + G(q) dx + dF(q) .
Now move G(q) dx to the left-hand side
[q − G(q)] dx = x dG(q) + dF(q)
and divide both sides by dq
dx dG(q) dF(q)
[q − G(q)] = x + .
dq dq dq
Assume q = G(q). Multiply from the left both sides by 1

[q − G(q)]
.
dx 1 dG(q) dF(q)
= x + , (6.20)
dq [q − G(q)] dq dq
This is a first-order differential equation, and its solution, x(q), should be achievable
depending on the details of the functions G(q) and F(q). Even its singular solution
may be found be setting the singular value q = G(q) in the original differential
equation.
6.2.2 Examples II(A)–II(C)

3q
I I (A) : y = x − q2 .
2
2
q q
I I (B) : y = x − .
3 6
I I (C) : y = 3 q x + 5 log(q) . (6.21)
Solution: II(A)
Here, G(q) = 3q
2
and F(q) = −q 2 . Therefore, (6.20) and II(A) give

dx 1 3
= −q x − 2q
dq 2

2
3
=− x +4 . (6.22)
q
This is a first-order differential equation. Its solution x(q) can be found by using
(4.10) and (4.15). The result is
σ1
x(q) = q + , (6.23)
q3
where σ1 is a constant. Inserting (6.23) into Lagrange equation II(A) leads to

1 2 3
y(q) = q + σ1 . (6.24)
2 q2
Equations (6.23) and (6.24) are parametric representation of the general solution
y = y(x).
Singular solution is obtained by setting q = G(q). That means using the relation-
ship q = 3 q2 , which is q = 0, in equation II(A). Thus, the singular solution is
y(x) = 0 . (6.25)
Solution: II(B)
q2
Here, G(q) = q3 and F(q) = − 6
. Therefore, (6.20) and II(B) give
dx 1 x q
= −
dq 2q 3 3
3
6.2 Lagrange Equation 129

1 x
= −1 . (6.26)
2 q
This is a first-order differential equation. Its solution, according to (4.10) and (4.15),
is
√
x(q) = − q − σ1 q . (6.27)
As usual, σ1 is an arbitrary constant. Inserting (6.27) into Lagrange equation II(B)

leads to
1 3 2 3
y(q) = − q + σ1 q 2 . (6.28)
3 2
Equations (6.27) and (6.28) are parametric equations representing the general solu-
tion y = y(x). Singular solution is obtained in similar fashion to that for equation
II(A). And the result once again is y(x) = 0.
Solution: II(C)
Here, G(q) = (3 q) and F(q) = 5 log(q). Therefore, (6.20) and II(C) give

dx 1 5
=− 3x + . (6.29)
dq 2q q
This is a first-order differential equation. Its solution, according to (4.10) and (4.15),
is

5 σ1
x(q) = − + 3 . (6.30)
q q2
As usual, σ1 is an arbitrary constant. Inserting (6.30) into Lagrange equation II(C)

leads to

σ1
y(q) = 5[log(q) − 3] + 3 1 . (6.31)
q2
Equations (6.30) and (6.31) are parametric equations representing the general solu-
tion y = y(x). Singular solution is obtained in similar fashion to that for equation
II(A). And the result once again is y(x) = 0.
Solve the following Lagrange differential equations problems labeled (1)–(3). Find
both the general solution and the singular solution. [Hint: Read (6.21)–(6.31).]
q
(1) : y = x − 3 q2 .
3
2
q q
(2) : y = x − .
5 5
(3) : y = 4 q x + 4 log(q) .
6.3 Separable Differential Equations
Whenever it is possible to separate the dependent and the independent variables and
express the differential equation in the form
dy dx
= (6.32)
Q(y) P(x)
the resultant equation

dy dx
= (6.33)
Q(y) P(x)
can directly be integrated.
Variables are separated in the following four problems, and then, direct integration
is used.
dy
(1) : = exp(x) dx .
y log y
dy
(2) : = exp(2 x) exp(−2 y) .
dx
d2 y dy
(3) : (sin x) 2 = (cos x) .
dx dx
d3 y
(4) : = (x 2 + 2x) exp x . (6.34)
dx 3
6.3 Separable Differential Equations 131
6.3.2 Solution

dy
(1) : = exp(x) dx ; log[log(y)] = exp(x) + const .
y log y

Equivalently : y = log−1 log−1 exp(x) + σ0 .

dy exp(2 y) exp(2 x)
(2) : = exp(2 x) dx ;
= + const .
exp(−2 y) 2 2
1
Equivalently : y = log σ0 + exp(2 x) .
2
dy d2 y dq dy
(3) : U se = q ; (sin x) 2 = (sin x) = (cos x) = (cos x) q
dx dx dx dx

dq
T her e f or e, = cot x dx ; log q = log(sin x) + const ;
q
dy
Equivalently , q ≡ = σ0 sin x .
dx

Or , dy = σ0 sin x dx ; y = − σ0 cos x + σ1 .

d3 y d2 y
(4) : dx = = (x 2 + 2x) exp (x) dx = x 2 exp x + σ0
dx 3 dx 2
2
d y dy
T her e f or e , dx = = x 2 . exp (x) dx + σ0 x
dx 2 dx
= (x 2 − 2x + 2) exp x + σ0 x + σ1 .

T hus , dy = [(x 2 − 2x + 2) exp x + σ0 x + σ1 ] dx .
σ0 2
Or , y = (x 2 − 4x + 6) exp x + x + σ1 x + σ2 . (6.35)
2
Solve the following problems by separating the variables.
dy dx
(1) : = .
y x
dy
(2) : = exp(3y − 2x) .
dx
dy
(3) : = 2 x y exp(x 2 ) .
dx
dy
(4) : = x 2 y2 .
dx
dy
6.4 Separable Equations of Form dx = xy
Another readily separable class of equations are of the form
dy y
= . (6.36)
dx x
dy
6.4.1 Solution dx = xy
Introduce a new variable

y
ρ= . (6.37)
x
As a result, y = x ρ and (6.36) can be represented as
dy d(xρ) dρ
(ρ) = = = x +ρ . (6.38)
dx dx dx
Transferring ρ to the left-hand side,
dρ
(ρ) − ρ = x ,
dx
dx 1
and multiplying from the left both sides by x (ρ)−ρ
finally transforms (6.36) into
one of separable form
dx dρ
= . (6.39)
x (ρ) − ρ
Equation (6.39), depending on complexity of (ρ), may now be integrated.

dy
6.4 Separable Equations of Form dx = xy 133
dy
6.4.2 Examples Group IV: dx = xy Equations
Examples Group IV: Equations (1)–(6)
dy y y 2
(1) : = 1+ + .
dx x x
dy y y
(2) : = tan + .
dx x x
dy y y
(3) : = exp − + .
dx x x
dy x y
(4) : = + .
dx y x
dy
(5) : x (x 2 + y 2 ) = y (x + y)2 .
dx
dy y + cx
(6) : = . (6.40)
dx x + cy
6.4.3 Solution
y
As noted in (6.38) and (6.37), substitute (ρ) for dy
dx
and ρ for x
. One gets
(1) : (ρ) = 1 + ρ + ρ2 . T her e f or e, accor ding (6.39)

dx dρ dρ
= = , or
x (ρ) − ρ 1 + ρ2

σ0 + log x = tan−1 ρ , or ρ = tan σ0 + log x , or

y = x tan σ0 + log x .

dρ dx
(2) : (ρ) = tan ρ + ρ . T her e f or e, dx = , or
tan ρ x
log(sin ρ) = log σ0 + log x , sin ρ = σ0 x , or
y = x sin−1 [σ0 x] .

dρ dx
(3) : (ρ) = exp(−ρ) + ρ . T her e f or e, = , or
exp(−ρ) x

exp(ρ) = σ0 + log x; ρ = log σ0 + log x , or y = x log σ0 + log x .

1 dρ dx
(4) : (ρ) = + ρ . T her e f or e, = , or
ρ 1
ρ
x
ρ2 √ 1
= σ0 + log x , or y = ± x 2 σ0 + log x 2 .
2

dy y 2 xy
(5) W rite as : = 1+ 2 .T her e f or e ,
dx x 1 + xy

2ρ2 1 + ρ2 dx
(ρ) = + ρ . Accor dingly, dρ = , or
1 + ρ2 2ρ2 x
1 ρ
− + = σ0 + log x . T hus : ρ2 − 2ρ (σ0 + log x) − 1 = 0 , or
2ρ 2

ρ = (σ0 + log x) ± (σ0 + log x)2 + 1 , or

y = x (σ0 + log x) ± x (σ0 + log x)2 + 1.
(6) : (ρ) = (ρ + c)/(1 + cρ) . T her e f or e,

dρ dρ 1 1 + cρ
= = dρ
(ρ) − ρ (ρ + c)/(1 + cρ) − ρ c 1 − ρ2

dρ dρ dx
= (2c)−1 −(c − 1) + (c + 1) =
1+ρ 1−ρ x

= − (2c)−1 (c − 1) log(1 + ρ) + (c + 1) log(1 − ρ) = σ0 + log x . (6.41)

dy y
Solve the following dx
= x
problems . [Hint: Read (6.40) and (6.41).]
dy y y 2
(1) : =2 + .
dx x x
dy y y
(2) : = 3 tan + .
dx x x
dy y y
(3) : = 5 exp −3 + .
dx x x
dy x y
(4) : =4 + .
dx y x
dy
(5) : x (x 2 + 2 y 2 ) = y x 2 + 3 y x + 2 y2 .
dx
dy y
6.5 Equations Reducible to dx = x 135
dy y
6.5 Equations Reducible to dx = x
These are equations of the form
dy b1 y + c1
= φ .
dx b2 x + c2
6.5.1 Examples Group V: (I)–(IV)
Solved below are equations (I ) to (I V ) that are of the form described above. Despite
slightly different appearance, they are readily reducible to the general equation dy =
dx
xy .
2
dy y y
(I ) : = 1+ + .
dx x +2 x +2
dy y+1
(I I ) : = 2 +4 .
dx x +2
dy y+2 y+2
(I I I ) : = tan + .
dx x +3 x +3

dy y+3 y+3
(I V ) : = exp − + .
dx x +4 x +4
Solution: (I)
dz dy dy dz dy
Set (x + 2) = z . T hen = 1 and = = ·1
dx dx dz dx dz
2
dy dy y y
T her e f or e , = = 1+ + . (6.42)
dz dx z z
Except for change of variable from x → z, this is identical to (1) in (6.40). Therefore,
its solution is

(I ) : y = z tan σ0 + log z , or , y = (x + 2) tan σ0 + log(x + 2) .
(6.43)
Solutions: (II)–(IV)
Set z 1 = y + 1 , z 2 = x + 2 and z1
z2
= ρ . Then
dy dz 1
= .
dx dz 2
Also, rewriting equation (II), one has

dy z1
= 2 + 4 = 2ρ + 4.
dx z2
If one sets
2ρ + 4 = (ρ) ,
then, much like (6.38), one has dy = (ρ). Therefore, according to (6.39),
dx
dx
x
= (ρ)−ρ . Additionally, because dy
dρ
dx
= dz
dz 1
2
, the earlier variable y corresponds
to the current variable z 1 . Similarly, the earlier variable x corresponds to the current
variable z 2 . Therefore, for brevity it is convenient to go directly to (6.39) and rewrite
it by transliterating x to z 2 . Additionally, we do the same for the corresponding result
that would be obtained.
dx dz 2 dρ dρ dρ
→ = = = . (6.44)
x z2 (ρ) − ρ 2ρ + 4 − ρ ρ+4
Integrating both sides leads to

dz 2 dρ
= ,
z2 ρ+4
log z 2 + constant = log(ρ + 4) ,
σ0 z 2 = ρ + 4 . (6.45)
Using the information, ρ = z1

z2
, z 1 = y + 1, z 2 = x + 2, the above relationship
becomes σ0 (x + 2) = y+1
x+2
+ 4 which leads to the result
(I I ) : y = σ0 (x + 2)2 − 4x − 9 . (6.46)

z1
(I I I ) : Set z 1 = y + 2 , z 2 = x + 3 , =ρ .
z2

dy dz 1 z1 z1
T hen , = = tan + = tan ρ + ρ . (6.47)
dx dz 2 z2 z2

As per (6.38) and (6.39), the relationship dxx = dρ
(ρ)−ρ
obtains. And again because
dy
here dx
= dz 1
dz 2
, therefore, similar to (6.39), one can write
dy y
dz 2 dρ dρ dρ
= = = . (6.48)
z2 (ρ) − ρ tan ρ + ρ − ρ tan ρ

dz 2 dρ
= ,
z2 tan ρ
log z 2 + constant = log(sin ρ) , (6.49)
σ0 z 2 = sin ρ ; ρ = sin−1 (σ0 z 2 ) . (6.50)
Using the information, z 1 = y + 2, z 2 = x + 3, ρ = z1

z2
, the above relationship
y+2 −1
becomes x+3
= sin [σ0 (x + 3)] which leads to the result
(I I I ) : y = (x + 3) sin−1 [σ0 (x + 3)] − 2 . (6.51)

z1
(I V ) : Set z 1 = y + 3 , z 2 = x + 4 , =ρ .
z2

dy dz 1 z1 z1
T hen , = = exp − + = exp(−ρ) + ρ . (6.52)
dx dz 2 z2 z2
Following the same procedure as used in the above two equations, namely (II) and
(III), one can write
dz 2 dρ dρ dρ
= = = . (6.53)
z2 (ρ) − ρ exp(−ρ) + ρ − ρ exp(−ρ)

dz 2 dρ
= ,
z2 exp(−ρ)
log(z 2 ) + σ0 = exp(ρ) ,

ρ = log log(z 2 ) + σ0 . (6.54)
Because z 1 = y + 3, z 2 = x + 4, ρ = z1
z2
, (6.54) readily leads to the result

(I V ) : y = (x + 4) log log(x + 4) + σ0 − 3 . (6.55)
dy a1 x+b1 y+c1
6.5.2 Equations dx = a2 x+b2 y+c2

dy x−y
(V ) : = .
dx x +y+2

dy x −y+1
(V I ) : = .
dx x +y+3
dy x + 2y + 3
(V I I ) : = . (6.56)
dx 4(x + 2y) + 5
Solutions of (V ) and (V I I )
Unlike (II)–(IV) the situation is different here because either the numerator or the
denominator, or both, contain x as well as y. When such is the case, one proceeds as
follows: First one sets

z1 dy dz 1
y = z1 + a ; x = z2 + b ; =ρ; = . (6.57)
z2 dx dz 2
Then, one eliminates both a and b as well as the original constants in the numerator
and the denominator. To that end, one employs Cramer’s rule—see (3.117). This
procedure is demonstrated in the treatment of (V ) and (V I ) below.
Equation (V): Solution

dy x−y z 2 − z 1 + (−a + b)
(V ) : = = . (6.58)
dx x +y+2 z 2 + z 1 + (a + b + 2)
Both (−a + b) and (a + b + 2) are vanishing when b = a = −1. As such z 1 =

y+1
(y − a) = (y + 1), z 2 = (x − b) = (x + 1), and ρ = zz21 = x+1 .
Accordingly, the differential equation (V ) becomes
dy dz 1 z2 − z1 1−ρ
= = = = (ρ) . (6.59)
dx dz 2 z1 + z2 1+ρ
Following the same procedure as in (I I )−(I V ), one can write

⎡ ⎤
dz 2 dρ dρ (1 + ρ)
= =⎣ ⎦ = dρ . (6.60)
z2 (ρ) − ρ 1−ρ
−ρ 1 − 2ρ − ρ2
1+ρ

dy y
1
log(z 2 ) = − log(ρ2 + 2ρ − 1) + σ1 . (6.61)
2
Replacing z 2 by (x + 1) and ρ by y+1

x+1
gives
1
log(x + 1) = − log[y 2 + 4y + 2yx − x 2 + 2] + log(x + 1) + σ1 .
2
Thus
log[y 2 + 4y + 2yx − x 2 + 2] = 2 σ1 .
Exponentiation leads to
y 2 + 4y + 2yx − x 2 − σ0 = 0 .
This is a quadratic in y with solutions:

(V ) : y = −(x + 2) ± 2x 2 + 4x + σ0 . (6.62)

dy x −y+1 z 2 − z 1 + (−a + b + 1)
(V I ) : = = . (6.63)
dx x +y+3 z 2 + z 1 + (a + b + 3)
Equation (VI) : Solution

Both (−a + b + 1) and (a + b + 3) are vanishing when a = −1 and b = −2. As
y+1
such z 1 = (y − a) = (y + 1), z 2 = (x − b) = (x + 2) and ρ = zz21 = x+2 . As a
result, the differential equation (V I ) becomes
dy dz 1 z2 − z1 1−ρ
= = = = (ρ) . (6.64)
dx dz 2 z1 + z2 1+ρ
In this form, it becomes identical to (V ) above—see (6.59). Therefore, its solution

is as obtained in (6.62).
1
log(z 2 ) = − log(ρ2 + 2ρ − 1) + σ1 . (6.65)
2
Replacing z 2 by (x + 2) and ρ by y+1

x+2
leads to
1
log(x + 2) = − log[y 2 + 6y + 2yx − x 2 − 2x + 1] + log(x + 2) + σ1 .
2
Thus
log[y 2 + 6y + 2yx − x 2 − 2x + 2] = 2 σ1 .
Exponentiating the above gives
y 2 + 6y + 2yx − x 2 − 2x − σ0 = 0
which is a quadratic in y with solutions:

(V I ) : y = −(x + 3) ± 2x 2 + 8x + σ0 . (6.66)
Remark
Differential equation (V I ) that was solved above had variables x, y as well as con-
stants both in the numerator and the denominator. It was successfully treated by
replacing x and y with two new variables such that the use of Cramer’s rule helped
eliminate the constants. There are, however, situations where the Cramer’s rule is
inapplicable. When that is the case, and the differential equation has features similar
to (V I I ), one proceeds somewhat differently.
Procedure for Solving (VII)
For generality and ease of recognizing the feature of interest, it is convenient to
rewrite problem (VII) as
x + 2y + 3
dy = dx
4(x + 2y) + 5
(a x + b y) + c
≡ dx (6.67)
j (a x + b y) + r
where a, b, c, j, and r are constants. Clearly, the Cramer’s rule is inapplicable here
because the relevant determinant,

a b

j a j b , (6.68)
is vanishing.
To deal with this matter, let us introduce a single variable, z 0 , for the combination
(ax + by).
z0 = a x + b y . (6.69)
Differentiate (6.69) and use (6.67) and (6.69).

dy y
dz 0 = a dx + b dy
(a x + b y) + c
= a dx + b dx
j (a x + b y) + r
z0 + c
= a dx + b dx
j z0 + r
= dx [z 0 (a j + b) + (a r + b c)] /( j z 0 + r ) . (6.70)
( j z 0 +r )
Next, multiply the resultant equation (6.70) by [z 0 (a j+b)+(a r +b c)]
.
j z0 + r
dz 0 = dx. (6.71)
z 0 (a j + b) + (a r + b c)
As a result, the left- and the right-hand sides are functions of only one variable each:
that is z 0 and x. Thus, one can directly integrate both sides to determine one as a
function of the other. Finally, because z 0 is equal to (a x + b y), in this manner one
has determined the desired solution y ≡ y(x).
Equation (VII) : Solution

Comparing (VII) and its surrogate equation (6.67), one finds the relationships: a →
1 ; b → 2 ; c → 3 ; j → 4 ; r → 5. Accordingly, (6.71) becomes
4 z0 + 5
dz 0 = dx . (6.72)
6 z 0 + 11
2 z0 7
− log(6 z 0 + 11) + σ0 = x .
3 18
Setting z 0 = (a x + b y) → (x + 2 y), one finds the requisite solution to (VII).
2(x + 2 y) 7
(V I I ) : x = − log (6 x + 12 y + 11) + σ0 . (6.73)
3 18
Solve the
following problems, labeled (1)–(7), that are in fact reducible to the form
q = xy . [Hint: Read (6.42)–(6.73).]
2
dy y y
(1) : = 1+ + .
dx x +1 x +1
dy y+2
(2) : = +1 .
dx x +3
dy y+1 y+1
(3) : = tan + .
dx x +2 x +2

dy y+2 y+2
(4) : = exp − + .
dx x +3 x +3

dy x −y+1
(5) : = .
dx x + 2y + 3

dy x −y+1
(6) : = .
dx x + 3y + 4
dy (x + y) + 2
(7) : = .
dx 3(x + y) + 4
6.6 Exact Differential and Exact Differential Equation
6.6.1 Exact Differential
Consider variables X and Y that are both real and finite within an(X,Y ) domain
,
and a function Z (X, Y ) that has continuous partial derivatives ∂∂ XZ Y and ∂∂YZ X .
Then, the total derivative of Z , namely dZ (X, Y ),
dZ (X, Y ) = U (X, Y ) dX + V (X, Y ) dY , (6.74)
is exact if U (X, Y ) and V (X, Y ) are functions of X and Y in the domain and obey
the so-called integrability requirement:

∂Z ∂Z
U (X, Y ) = ; V (X, Y ) = . (6.75)
∂X Y ∂Y X
This requirement holds if the second mixed derivatives of Z are equal. That is, if the
following is true.
∂2 Z ∂2 Z
= . (6.76)
∂Y ∂ X ∂ X ∂Y
Thus, in two dimensions, the standard representation for an exact differential is
6.6 Exact Differential and Exact Differential Equation 143

∂Z ∂Z
dZ (X, Y ) = dX + dY (6.77)
∂X Y ∂Y X
with an important caveat that the following equality holds.

∂Z ∂Z
∂ ∂X Y ∂2 Z ∂2 Z ∂ ∂Y X
≡ = ≡ . (6.78)
∂Y ∂Y ∂ X ∂ X ∂Y ∂X
X Y
6.6.2 Exact Differential Equation
Given an exact differential dZ (X, Y ), the relationship
dZ (X, Y ) = 0 , (6.79)
or equivalently

∂Z ∂Z
dX + dY = 0 , (6.80)
∂X Y ∂Y X
is referred to as an exact differential equation.
6.6.3 Requirement
Consider differential equation (6.81).
dZ (X, Y ) = U (X, Y ) dX + V (X, Y ) dY

=0 . (6.81)
Its exactness requires (6.82) to be satisfied.

∂Z ∂Z ∂U ∂V
= U (X, Y ) ; = V (X, Y ) ; = .
∂X Y ∂Y X ∂Y X ∂X Y
(6.82)
6.6.4 Solution
Consider differential equation (6.81)
dZ (X, Y ) = 0 (6.83)
and assume it is exact. Solving it requires the evaluation of the function Z (X, Y ). To
that purpose, proceed as follows: One has from (6.81) and (6.82):

∂Z
U (X, Y ) = . (6.84)
∂X Y
Integrating both sides with respect to X gives

∂Z
U (X, Y ) dX = dX = Z (X, Y ) + F(Y ) . (6.85)
∂X Y
The function F(Y ) is as yet undetermined.

To eliminate it, proceed as follows.
Equation (6.85) relates F(Y ) to U (X, Y ) dX.
Symmetry considerations require the next task must be to relate F(Y ) to
V (X, Y ) dY.
In order to do that, one needs to differentiate (6.85) with respect to Y while holding
X constant.

∂[ U (X, Y ) dX ] ∂Z dF(Y )
= +
∂Y X ∂Y X dY
dF(Y )
= V (X, Y ) + . (6.86)
dY

In the above, following (6.82), ∂∂YZ X was replaced by V (X, Y ). This establishes
the needed connection. And F(Y ) can now be determined by integrating both sides
of (6.86) with respect to Y.

∂[ U (X, Y ) dX ] dF(Y )
dY = V (X, Y ) dY + dY + σ
∂Y dY
X

= V (X, Y ) dY + F(Y ) + σ . (6.87)
Equations (6.85) and (6.87) are rewritten as

−F(Y ) = − U (X, Y ) dX + Z (X, Y ) ;

∂[ U (X, Y ) dX ]
−F(Y ) = V (X, Y ) dY − dY + σ .
∂Y X
Eliminate −F(Y ). Thus, the solution Z (X, Y ) of the exact differential equa-
tion (6.81) is

Z (X, Y ) = U (X, Y ) dX + V (X, Y ) dY

∂[ U (X, Y ) dX ]
− dY + σ . (6.88)
∂Y X
Given below are six differential equations that are similar to (6.81).
(1) : 0 = dZ = [1 + 2Y ] dX + [2 + 2X ] dY .

(2) : 0 = dZ = 3X 2 + 4X Y dX + 2X 2 + 2Y dY .

(3) : 0 = dZ = exp(2Y ) dX + 2X exp(2Y ) − Y dY .

(4) : 0 = dZ = 3Y X 2 + sin(Y ) dX + X cos(Y ) + X 3 + 1 dY .

(5) : 0 = dZ = sin(Y ) + X Y 2 + X 2 Y + X + Y dX
X3
+ + X cos(Y ) + X 2 Y + X dY .
3

(6) : 0 = dZ = X + 2X cos(Y ) + 3X 2 Y dX

+ X 3 − X 2 sin(Y ) + Y dY . (6.89)
6.6.6 Solution
Choose U (X, Y ) and V (X, Y ) from one of the six equations (6.89). Check whether
the relevant differential equation is exact by using (6.82). And if so, employ (6.88)
to find its solution.
The results are given below ad seriatim.
(1) : U = 1 + 2Y ; V = 2 + 2X .

∂U ∂V
= 2 = .
∂Y X ∂X Y

U dX = (X + 2Y X ) , V dY = (2Y + 2X Y ) ,

∂[ U dX ]
− dY = − 2X dY = − 2X Y ,
∂Y X
σ1 = (X + 2Y X ) + (2Y + 2X Y ) − (2X Y )

X − σ1
T her e f or e , Y = − .
2 + 2X
(2) : U = 3X 2 + 4X Y ; V = 2X 2 + 2Y .

∂U ∂V
= 4X = .
∂Y X ∂X Y
σ1 = (X 3 + 2Y X 2 ) + (Y 2 + 2X 2 Y ) − (2X 2 Y )
T her e f or e , Y 2 + 2X 2 Y + X 3 − σ1 = 0 .

Or equivalently , Y = − X 2 ± X 4 − X 3 + σ1 .
(3) : U = exp(2Y ) ; V = 2X exp(2Y ) − Y.

∂U ∂V
= 2 exp(2Y ) = .
∂Y X ∂X Y

U dX = X exp(2Y ) , V dY = X exp(2Y ) − Y 2 /2,

∂[ U dX ]
− dY = − 2X exp(2Y ) dY = − X exp(2Y ),
∂Y X
σ1 = X exp(2Y ) − Y 2 /2
T her e f or e , X = (Y /2) exp(−2Y ) + (σ1 ) exp(−2Y ) .
2
(4) : U = 3Y X 2 + sin(Y ) ; V = X cos(Y ) + X 3 + 1 .

∂U ∂V
= 3X + cos(Y ) =
2
.
∂Y X ∂X Y

σ1 = X sin(Y ) + Y X 3 + X 3 Y + X sin(Y ) + Y

− X sin(Y ) + Y X 3
T her e f or e , σ1 = X 3 Y + X sin(Y ) + Y .
(5) : U = sin(Y ) + X Y 2 + X 2 Y + X + Y ;
X3
V = + X cos(Y ) + X 2 Y + X .
3
∂U ∂V
= cos(Y ) + 2X Y + X + 1 =
2
.
∂Y X ∂X Y

σ1 = X sin(Y ) + X 2 Y 2 /2 + X 3 Y/3 + X 2 /2 + X Y
+ X 3 Y/3 + X sin(Y ) + X 2 Y 2 /2 + X Y

− X sin(Y ) + X 2 Y 2 /2 + X 3 Y/3 + X Y
T her e f or e , σ1 = X 3 Y/3 + X sin(Y ) + X 2 Y 2 /2 + X Y + X 2 /2 .
(6) : U = 2X cos(Y ) + 3X 2 Y + X ; V = X 3 − X 2 sin(Y ) + Y .

∂U ∂U
= 3 X 2 − 2 X sin(Y ) = .
∂Y X ∂Y X

σ1 = X 2 cos(Y ) + X 3 Y + X 2 /2 + X 3 Y + X 2 cos(Y ) + Y 2 /2

− X 2 cos(Y ) + X 3 Y .
T her e f or e , σ1 = X 3 Y + X 2 cos(Y ) + X 2 /2 + Y 2 /2 . (6.90)
Exercise
Due to symmetry, it is clear that the functions U (X, Y ) and V (X, Y ) may be inter-
changed as long as the same is done for the variables X and Y. Therefore, (6.88) can
also be represented as

Z (X, Y ) = V (X, Y ) dY + U (X, Y ) dX

∂[ V (X, Y ) dY ]
− dX + σ . (6.91)
∂X Y
Show that this is true. Also, by using (6.91)—that is, instead of (6.88)—solve the six
differential equations given as (6.89).
6.6.7 Problems Group VI
Solve the exact differential equations specified in problems (1) → (3). [Hint: Read
(6.88) and (6.90).]

(1) : 0 = dz = [1 + 2x y] dx + 2 + x 2 dy .

(2) : 0 = dz = 1 + y + y 2 dx + [2 + 2x y] dy .

(3) : 0 = dz = exp(y) + x dx + x exp(y) + y dy .
6.7 Inexact Differential Equation Integrating Factor
A differential equation of the form
u(x, y) dx + v(x, y) dy = 0 , (6.92)
that is not exact, meaning


∂u ∂v
= , (6.93)
∂y x ∂x y
may be made exact if multiplied by an appropriate function I (x, y). When that can
be done, and the new differential equation
I (x, y) [u(x, y) dx + v(x, y) dy] = 0 , (6.94)
is exact, the multiplying function I (x, y) is called an integrating factor.

Although, in principle, for every inexact differential equation there exists an appro-
priate integrating factor, in practice such integrating factors are hard to find. More-
over, a general rule for determining the relevant integrating factor is not available.
However, if the integrating factor should turn out to be a function of only a single
variable, progress can be made.
6.7.1 Integrating Factor Dependent only on x
Assume the integrating factor for the inexact differential equation (6.92) is dependent
only on x. That is
I (x, y) ≡ X (x) . (6.95)
In other words, the following differential equation is assumed to be exact
X (x).[u dx + v dy] = 0 . (6.96)
Clearly, if (6.96) is indeed exact, it must satisfy the exactness rule

∂[X (x).u] ∂[X (x).v]
= . (6.97)
∂y x ∂x y
That is

∂u ∂v dX (x)
X (x) = X (x) + v(x, y) . (6.98)
∂y x ∂x y dx
This readily leads to the relationship
∂u
∂v
∂y
− ∂x y
dX (x)
x
= dx
. (6.99)
v(x, y) X (x)
In other words, if the integrating factor X that makes the inexact differential equa-
tion (6.92) exact is to depend only on x, it must satisfy (6.99).
6.7 Inexact Differential Equation Integrating Factor 149
Multiplying both sides of (6.99) by dx and integrating gives

∂v
∂u
−
∂y ∂x y
x
dx = log X (x) + constant . (6.100)
v(x, y)
To conclude, if there is an integrating factor X (x) that makes an inexact differential

equation u(x, y) dx + v(x, y) dy = 0 exact, it must have the form
⎡ ⎤
∂u(x,y) ∂v(x,y)
∂y
− ∂x
⎢ x y ⎥
X (x) = σ1 exp ⎣ dx ⎦ . (6.101)
v(x, y)
6.7.2 Integrating Factor Dependent only on y
Assume there is an integrating function, Y (y), dependent only on y that makes the
inexact differential equation (6.92) exact. That is, assume the following is an exact
differential equation
Y (y) [u(x, y) dx + v(x, y) dy] = 0 . (6.102)
Comparing with (6.99) and (6.101), symmetry considerations suggest that if there is
an integrating factor Y (y) that makes an inexact differential equation u dx + v dy = 0
exact, it must have the form
∂v ∂u
∂x y
− ∂y
dY (y)
dy
= x
;
u(x, y) Y (y)
⎡ ⎤
∂v(x,y) ∂u(x,y)
∂x
− ∂y
⎢ y x ⎥
Y (y) = σ2 exp ⎣ dy ⎦ . (6.103)
u(x, y)
Exercise
Prove that (6.103), guessed on symmetry grounds, are correct.
Exact or Inexact?
Solve the following three differential equations.
(1) : 0 = dz = [1 + 2y] dx + [1 + 3x] dy .

(2) : 0 = dz = y 2 dx + 3x y dy .

(3) : 0 = dz = x 2 − y dx + x dy . (6.104)
Procedure: First choose u(x, y) and v(x, y) from one of the three equations (6.104).
Then by using (6.82), check as to whether the relevant differential equation is exact
or inexact. And if it is inexact, employ (6.99) or (6.103) to find whether an integrating
factor that depends only upon a single variable is possible.
Solution of (6.104)-(1)

∂u ∂v
u = 1 + 2y ; v = 1 + 3x . = 2 and = 3 . (6.105)
∂y x ∂x y
Because 2 = 3, this equation is inexact. Therefore, try using (6.99) to check as to

whether an integrating factor that depends only on the variable x is possible.
∂v
∂u
∂y
− ∂x y −dx dX (x)
x
.dx = = . (6.106)
v 1 + 3x X (x)
Good news! There exists an integrating factor X (x). One can find it by integrating
the above equation.
−1
log[X (x)] = log (1 + 3x) 3 + constant. (6.107)
Thus
σ0
X (x) = 1 . (6.108)
(1 + 3x) 3
Including the integrating factor, and ignoring the unnecessary multiplier σ0 , the
differential equation (6.104)-(1) now is
(1 + 3x)− 3 [(1 + 2y) dx + (1 + 3x) dy] = 0 .

1
(6.109)
Now, determine whether this new differential equation is indeed exact. One has
u(x, y) = (1 + 2y)(1 + 3x)− 3 ; v(x, y) = (1 + 3x) 3 . T her e f or e,

1 2

∂u ∂v 2
= 2(1 + 3x)− 3 ; (1 + 3x)− 3 3 . (6.110)
1 1
=
∂y x ∂x u 3
Good! The new equation is exact. Accordingly, one can find its solution via the
procedure outlined in (6.88). That is

∂[ u(x, y) dx]
σ1 − σ2 = u(x, y) dx +v(x, y) dy − dy
∂y
x
= (1 + 3x)− 3 (1 + 2y)dx + (1 + 3x) 3 dy

1 2

∂[ u(x, y) dx]
− dy
∂y x
2
(1 + 2y)(1 + 3x) 3 2 2
= 2 + (1 + 3x) 3 y − (1 + 3x) 3 y
3 3
2
(1 + 2y)(1 + 3x) 3
= . (6.111)
2
Therefore, the solution to (6.104)-(1) is
−1 σ1 − σ2
y(x) = + 2 . (6.112)
2 (1 + 3x) 3

∂u ∂v
u = y 2 ; v = 3x y ; = 2y = = 3y . (6.113)
∂y x ∂x y
This equation is clearly NOT exact. However, to check as to whether an integrating

factor that depends only on the variable y is possible, use (6.103).

∂v ∂u
∂x y
− ∂y
dY
dy
x
.dy = x
.dy
u Y
3y − 2y dY (y)
.dy = . (6.114)
y2 Y (y)
Good news! There exists an integrating factor Y (y), and one can find it by integrating
the above equation.
log(y) = log Y (y) + constant. (6.115)
Thus
Y (y) = σ0 y . (6.116)
With this integrating factor, the differential equation (6.104)-(2) becomes
y 3 dx + 3x y 2 dy = 0 . (6.117)
As before one checks as to whether this new differential equation is indeed exact.
u(x, y) = y 3 ; v(x, y) = 3y 2 x .

∂u ∂v
And, = 3 y2 ; = 3 y2 . (6.118)
∂y x ∂x u
Yes, the new equation is exact. Accordingly, one can find its solution via the procedure
outlined in (6.91). That is

∂[ v(x, y) dy]
σ1 − σ2 = v(x, y) dy + u(x, y) dx − dx
∂x y

∂[ v(x, y) dy]
= (3y x) dy + (y ) dx −
2 3
dx
∂x y
= y3 x + y3 x − y3 x = y3 x . (6.119)
Accordingly, the solution to (6.104)-(2) is

1
σ1 − σ2 3
y(x) = . (6.120)
x

∂u ∂v
u = x −y ; v = x ;
2
= − 1 = = 1 . (6.121)
∂y x ∂x y
This equation is NOT exact. Therefore, try (6.99) to check as to whether an integrating
factor that depends only on one variable is possible. To that end, x would appear to
be the more likely candidate.
∂v
∂u
∂y
− ∂x y −2 dx dX
x
.dx = = . (6.122)
v x X
Thus, there appears to be an integrating factor X that depends only on x. And it

obeys the relationship
log(X ) = − 2 log x + constant (6.123)
leading to
σ0
X (x) = . (6.124)
x2
With the integrating factor included, the differential equation (6.104)-(3) reads
1 2
2
x − y dx + x dy = 0 ;
x
1 1 1
u(x, y) = 2 x 2 − y ; v(x, y) = 2 x = . (6.125)
x x x
For exactness, it must satisfy the equality

∂u 1 ∂v
= − 2 = (6.126)
∂y x x ∂x u
which it does. Accordingly, one can find its solution via the procedure outlined
in (6.88). That is

∂[ u(x, y) dx]
σ1 − σ2 = u(x, y) dx + v(x, y) dy − dy
∂y x
y y y y
= x+ + − = x+ . (6.127)
x x x x
y(x) = (σ1 − σ2 )x − x 2 . (6.128)
6.7.4 Problems Group VII

[Hint: Compare (6.81)–(6.128).]
(1) : 0 = dz = (2 + 3y) dx + (3 + 2x) dy .
(2) : 0 = dz = y 3 dx + 2x y 2 dy .

1
(3) : 0 = dz = x 2 + y dx + dy. (6.129)
3x
6.8 Riccati Equation
A nonlinear differential equation of the form
dy
+ y 2 + f 1 (x) y + f 2 (x) = 0 (6.130)
dx
is sometime called a Riccati equation.
6.8.1 Treatment
Set
1 dz
= y . (6.131)
z dx
As a result,
dz
= zy. (6.132)
dx
Its differentiation gives
d2 z dy dz dy
2
= z +y =z + y2 z . (6.133)
dx dx dx dx
Now, multiply (6.130) by z
dy
z + y 2 + f 1 (x) y + f 2 (x) = 0 (6.134)
dx
dy
and replace z dx
according to (6.133). This gives
d2 z
− y 2 z + z y 2 + f 1 (x) z y + f 2 (x) z = 0 . (6.135)
dx 2
Remove −y 2 z + z y 2 and replace z y by dz

dx
. As a result, the Riccati equation (6.130)
is transformed into
d2 z dz
2
+ f 1 (x) + f 2 (x)z = 0 . (6.136)
dx dx
Notice that this differential equation, unlike the original Riccati equation, is linear.
But the removal of nonlinearity has resulted in increasing the order. And we now
have a second-order differential equation. But, because the new equation is linear,
there is a greater possibility of making progress. In particular, if f 1 (x) and f 2 (x)
should both turn out to be constants, (6.136) can be solved much like equations of
the type (3.1), (3.5), etc. And once z has been determined, y can be calculated by a
simple differentiation as is implicit in the form of (6.131).
6.8 Riccati Equation 155
First convert the Riccati equations given below into second-order linear equations
and then solve them.
dy
(1) : + y 2 + x y = 0.
dx
dy
(2) : + y 2 + x 2 y = 0.
dx
dy
(3) : + y 2 + y + 1 = 0.
dx
dy
(4) : + y 2 + 2 y + 2 = 0. (6.137)
dx
Converting the Riccati equations into second-order linODE’s is straightforward. All
one needs to do is use (6.130) and (6.136). Accordingly, (6.137) get transformed as
follows:
(1) → (1 ) : d2 z
dx 2
+ x dxdz
= 0
d2 z
(2) → (2 ) : dx 2
+ x dx =
2 dz
0
(3) → (3 ) :
2
d z
dx 2
+ dx dz
+z = 0
2
(4) → (4 ) : d z
dx 2
+ 2 dx + 2 z =
dz
0 (6.138)
6.8.3 Solution
Equations (6.137)-(1) and (6.138)-(1’)

Introduce the notation p = dz
dx
and write (6.250)-(1’) as
dp
+x p = 0 . (6.139)
dx
Equation (6.139) is readily solved. (See, for instance, (4.2), (4.10), and (4.15) that
describe how such first-order linODEs are solved.) That is
2
x
p = σ0 exp − . (6.140)
2
Because p = dz
dx
, its integration leads to z.
2
dz x
dx = z = p dx = σ0 exp − dx + σ1 . (6.141)
dx 2
And once z is known, y can be determined by using (6.131). That is
1 dz p σ0 exp − x2
y= = =
z dx z 2
σ0 exp − x2 dx + σ1
2
exp − x2
= 2
. (6.142)
exp − x2 dx + σ2
Equations (6.137)-(2) and (6.138)-(2’)

Follow the same procedure as in (6.139)–(6.142) and write (6.138)-(2’) as
dp
+ x 2 p = 0. (6.143)
dx
Again with the help of (4.2), (4.10), and (4.15) one gets
3
x
p = σ0 exp − . (6.144)
3
Because p = dz
dx
its integration leads to z.
3
x
dz = z = σ0 exp − dx + σ1 . (6.145)
3
And once z is known, y can be determined by using (6.131). Much like (6.142) one
can write y as
3
exp − x3
y= 3
. (6.146)
exp − x3 dx + σ2
Equations (6.137)-(3) and (6.138)-(3’)

Notice that (6.138)-(3’) is homogeneous linear ordinary differential equation with
constant coefficients.
And according to (3.13), the complimentary solution of a second-order homoge-
neous linear ordinary differential equation with constant coefficients can be written
as
z ≡ Scomp (x) = σ1 exp (k1 x) + σ2 exp (k2 x) . (6.147)

6.8 Riccati Equation 157
Here, k = k1 and k = k2 are solutions of the Characteristic Equation(E ch )

2
dz
for (6.138)-(3’). The procedure for finding E ch is to replace z by 1, dx by k, and ddz z2
by k 2 . Thus, the Scomp (x) for (6.138)-(3’) is found by solving the following E ch .
k2 + k + 1 = 0 (6.148)
which leads to
√
−1 + i 3
k = k1 = r + im = ,
2 √
−1 − i 3
k = k2 = r − im = . (6.149)
2
Therefore, using (6.147)–(6.149), the result for z, according to (3.21) and (3.22), can
be written as
z = σ1 exp (r − i m) x + σ2 exp (r + i m)x

= exp (r x) [σ3 sin(m x) + σ4 cos(m x)]
√ √
x 3x 3x
= exp − σ3 sin + σ4 cos . (6.150)
2 2 2
Or, alternatively, as
√
x 3
z = σ0 exp − cos σ5 − x . (6.151)
2 2
Note, one went from (6.150)–(6.151) by introducing two new arbitrary constants σ0
and σ5 such that σ3 = σ0 sin(σ5 ) and σ4 = σ0 cos(σ5 ). This is perfectly alright as
long as σ02 = σ32 + σ42 . And because all sigmas are arbitrary constants, this equality
is trivially achieved.
Given z—see, (6.151)—the solution y is readily found.
√ √
1 dz 1 3 3
y = = − + tan σ5 − x . (6.152)
z dx 2 2 2
Equations (6.137)-(4) and (6.138)-(4’)

Again follow a similar procedure to that used for (6.147)–(6.152) and write the E ch
for (6.138)-(4’) as
k 2 + 2k + 2 = 0 (6.153)
which leads to
k = k1 = r + im = − 1 + i ,
k = k2 = r − im = − 1 − i . (6.154)
Therefore,
z = exp (−x) [σ3 sin x + σ4 cos x]

≡ σ0 exp (−x) cos (σ5 − x) . (6.155)
And the solution y is
1 dz
y = = − 1 + tan (σ5 − x) . (6.156)
z dx
6.8.4 Problems Group VIII
Solve the following two problems that involve Riccati equations.

[Hint: Read (6.137)–(6.156).]
dy
(1) : + y 2 + 2 y + 1 = 0.
dx
dy
(2) : + y 2 + y + 2 = 0.
dx
6.9 Euler Equation
Among Euler’s multifarious contributions to mathematics, in particular to the theory

of differential equations, is his work related to linear ordinary differential equations
with constant coefficient. He showed how linear ordinary differential equations with
variable coefficients of the form, cn x n , given below
ν
"
cn x n D n u = B(x) , (6.157)
n=0
could be transformed into linear ordinary differential equations with constant coeffi-
cients. The latter are much easier to solve as was demonstrated in detail in Chaps. 3
and 4.
6.9 Euler Equation 159
The Euler procedure consists in arranging an appropriate change of the indepen-

dent variable that transforms the operator D into a new operator . One proceeds as
follows. Set
x = exp(t)
and differentiate with respect to x.
dt
1 = exp(t) . . (6.158)
dx
Consider an arbitrary function U (x, t) that is differentiable with respect to x and t.

Multiply both sides of (6.158) from the right by d Ud(x,t)
t
.
d U (x, t) dt dU (x, t)
1· = exp(t) . .
dt dx dt
d U (x, t)
= exp(t) ·
dx
d U (x, t) d U (x, t)
= =x . (6.159)
dt dx
Because U (x, t) is arbitrary, (6.159) implies x d

dx
= d
dt
: meaning
x D = . (6.160)
Next, let us look at x D (x D).
x D (x D) = x (D x) D + x 2 D 2
= () = x D + x 2 D 2 = + x 2 D 2 . (6.161)
Therefore
x 2 D 2 ≡ ( − 1) . (6.162)
Similarly

x D x 2 D 2 = [ ( − 1)] . (6.163)
One also has

x D x 2 D2 = x 2 x D2 + x 3 D3 = 2 x 2 D2 + x 3 D3
= 2 ( − 1) + x 3 D 3 . (6.164)
The left-hand sides of (6.163) and (6.164) are equal. The equality of their right-hand
sides yields
2 ( − 1) + x 3 D 3 = [ ( − 1)] , (6.165)
or
x 3 D 3 ≡ ( − 1)( − 2) . (6.166)
In order to proceed to the next order, multiply x 3 D 3 from the left by x D and
use (6.166) to appropriately replace one of the expressions equal to 3 x 3 D 3 .

x D x 3 D3 = x 3 x 2 D3 + x 4 D4 = 3 x 3 D3 + x 4 D4
= 3 ( − 1) ( − 2) + x 4 D 4 . (6.167)
Multiply the left-hand side of (6.166) by x D and the right-hand side by the same
amount, that is . One gets

x D x 3 D 3 = [( − 1)( − 2)] . (6.168)
Now that the left-hand sides of (6.167) and (6.168) are the same, one can claim
equality of their right-hand sides. That is
3 ( − 1) ( − 2) + x 4 D 4 = [( − 1)( − 2)] ,
which leads to the result
x 4 D 4 ≡ ( − 1)( − 2)( − 3) . (6.169)
Exercise
Clearly, (6.160), (6.162), (6.166), and (6.169) show a pattern. Therefore, by induc-
tion, one assumes
x n D n = ( − 1)...( − n + 1) . (6.170)
Assuming (6.170) is valid when n = − 1, show that it is also valid when n = .
6.9.1 Examples Group IX
Transform each of the four differential equations I X − (A) → I X − (D) given

below by using the transformation x = exp(t), and renaming the function u(x) as
U (t). Finally, represent the result in terms of x. Then, find u(x).

I X − (A) : x 2 D 2 + x D + 1 u(x) = a x 2 + b . (6.171)
Solution IX-(A)
Use (6.158), (6.160), and (6.162) and transform (6.171) into
2
[( − 1) + + 1] U (t) = + 1 U (t) = a exp(2 t) + b . (6.172)
Then, solve for U (t) by following the familiar routine.
E ch = k 2 + 1 = 0 ; k1,2 = ± i ;
Scomp (t) = σ0 exp(−i t) + σi exp(−i t) = σ2 sin(t) + σ3 cos(t) ;
1 1 b
I pi (t) = 2 {a exp(2 t) + b} = a exp(2 t) 2 + ;
+1 2 +1 1
a
U (t) = Scomp (t) + I pi (t) = σ2 sin(t) + σ3 cos(t) + exp(2 t) + b .
5
Finally transform U (t) back into u(x). In this fashion, one gets:
a 2
U (t) → u(x) = σ2 sin[log(x)] + σ3 cos[log(x)] + x + b. (6.173)
5
Solution IX-(B)

I X − (B) : x 2 D 2 − 2 x D + 4 u = a log(x) . (6.174)
As above use (6.158), (6.160), and (6.162) and transform (6.174) into the following
differential equation.
2
− 3 + 4 U (t) = a t . (6.175)
Its solution is found in the usual manner.

√
3 7
E ch = k 2 − 3 k + 4 = 0 ; k1,2 = ± i ;
2 2

√ √
3t 7 3t 7
Scomp (t) = σ1 exp sin t + σ2 exp cos t ;
2 2 2 2

1 1 3 a 3a
I pi (t) = 2 (a t) = + (a t) = t + ;
− 3 + 4 4 16 4 16
U (t) = Scomp (t) + I pi (t) ;
√ √
7 3 7 3
U (t) → u(x) = σ1 sin log(x) x + σ2 cos
2 log(x) x 2
2 2
a 3a
+ log(x) + . (6.176)
4 16
Solution IX-(C)

I X − (C) : x 2 D 2 + 2x D − 4 u = a x 2 log(x) + b . (6.177)
Again use (6.158), (6.160), and (6.162) and transform (6.177) into
2
+ − 4 U (t) = a exp(2 t) t + b , (6.178)
Its solution is as follows:

√
1 17
E ch = k 2 + k − 4 = 0
; k1,2 = − ± ;
2 2
√ √
t 17 t 17
Scomp (t) = σ1 exp − + t + σ2 exp − − t ;
2 2 2 2
1 1
I pi (t) = 2 {a exp(2 t) t + b} = a exp(2t) t
+−4 { + 2) + ( + 2) − 4}
2

b 1 b 1 5 b
+ = a exp(2t) t − = a exp(2t) − t− ;
−4 2 + 5 + 2 4 2 4 4
1 5 b
I pi (t) = a exp(2t) t − − ;
2 4 4
U (t) = Scomp (t) + I pi (t);
√ √
1 17 1 17
U (t) → u(x) = σ1 x − 2 + 2 + σ2 x − 2 − 2
1 5
+ a x2 log(x) − . (6.179)
2 4
Solution IX-(D)

I X − (D) : x 2 D 2 − 4x D + 2 u = 2 x cos log(x) . (6.180)
And finally, as before, use (6.158), (6.160), and (6.162) and transform (6.180) into

2 − 5 + 2 U (t) = 2 exp(t) cos(t) . (6.181)
This leads to
√
5 17
E ch = k 2 − 5 k + 2 ; k1,2 = ± ;
2 2
√ √
5 17 5 17
Scomp (t) = σ1 exp t+ t + σ2 exp t− t ;
2 2 2 2
1 1
I pi (t) = 2 {2 a exp(t) cos(t)} = exp[t (1 + i)]
− 5 + 2 2 − 5 + 2
1
+ 2 exp[t (1 − i)];
− 5 + 2
1 1
= exp[t (1 + i)] + exp[t (1 − i)]
−3 − 3i −3 + 3i
exp(t) 1
=− (1 − i) exp(i t) + (1 + i) exp(−i t) = I pi (t) = − exp(t) [sin(t) + cos(t)] .
6 3
U (t) = Scomp (t) + I pi (t)

√ √
U (t) → u(x) = σ1 x 2 + 2 + σ2 x 2 −
5 17 5 17
2
x
− sin{log(x)} + cos{log(x)} . (6.182)
3
6.9.2 Euler Equation: An Extension
A simple extension of Euler equation is given below in (6.183).

ν
"
cn (a x + b)n D n u = B(x), (6.183)
n=0
Any such linear ordinary differential equation with variable coefficients of the form
cn (a x + b)n can readily be transformed into a linear ordinary differential equation
with constant coefficients—meaning into equations of the type that we have been
studying previously. The relevant transformation is carried out by a change of the
variable: from (a x + b) to exp(a t) where both a and b are constants. That is
(a x + b) = exp(a t). (6.184)
Differentiate (6.184) with respect to x, and divide both sides by the constant a.
dt
1 = exp(a t) .
dx
dt
= (a x + b) . . (6.185)
dx

Multiply both sides of (6.185) from the right by dUdt(x,t) .
dU (x, t) dt dU (x, t)
= (a x + b) .
dt dx dt
dU (x, t)
= (a x + b) . . (6.186)
dx
Because U (x, t) is arbitrary, this implies d

dt
= (a x + b) dxd : meaning
= (a x + b) D. (6.187)
Next, let us look at (a x + b) D {(a x + b) D} or in other words at {}.
(a x + b) D {(a x + b) D} = (a x + b)[a D + (a x + b) D 2 ]
= {} = a + (a x + b)2 D 2 . (6.188)
Transferring a across to the left side gives
(a x + b)2 D 2 = ( − a) . (6.189)
Similarly

(a x + b) D (a x + b)2 D 2 = (a x + b) 2a 2 x + 2ab D 2 + (a x + b)[(a x + b)2 D 3 ]
= 2a(ax + b)2 D 2 + (ax + b)3 D 3

= 2 a ( − a) + (a x + b)3 D 3 . (6.190)
By using (6.189) and the equivalence (a x + b) D = on the left-hand side of

(6.190), one gets
[ ( − a)] = 2 a ( − a) + (a x + b)3 D 3 , (6.191)
which leads to
(a x + b)3 D 3 = ( − a)( − 2 a) . (6.192)

Again, starting with (a x + b) D (a x + b)3 D 3 , one can readily show that
(a x + b)4 D 4 = ( − a)( − 2 a)( − 3 a) . (6.193)
Also, in view of (6.187), (6.189), (6.192), (6.193), or equivalently (6.194) given

below, one can transform a general equation of the type (6.183) into one involving
’s and functions of the variable t. One must not forget, however, to use (ax + b) =
exp(at) for transforming out of the variable B(x) into an appropriate function of the
variable t. This will become clear in examples (X I ) given below.
Exercise
Clearly, (6.187), (6.189), (6.192), and (6.193) show a pattern. Therefore, by induction
(a x + b)n D n = { − a}...{ − (n − 1)a} . (6.194)
6.9.3 Examples Group X
Extended Euler Equation

Transform each of the following three differential equations by setting the variable
(a x + b) = exp(a t) and then find their solution U (t). Also write the corresponding
solution u(x).

X (A) : (2x + 3)2 D 2 + 2(2x + 3) D − 4 u(x) = 5 log[2x + 3] . (6.195)
Solution of Example X(A)

Set (2x + 3) = exp(2 t) and (2x + 3)D = , use (6.187) and (6.189), and trans-
form (6.195) into the following:

[( − 2) + 2 − 4] U (t) = 5 log exp(2t)

= 2 − 4 U (t) = 10 t . (6.196)
Now proceed with the usual routine.
E ch = k 2 − 4 = 0 ; k1,2 = ± 2 ;
Scomp (t) = σ0 exp(2 t) + σ1 exp(−2 t) ;
1 1 2
I pi (t) = 2 {10 t} = − 1 + {10 t}
−4 4 4
10
=− t ;
4
10
U (t) = Scomp (t) + I pi (t) = σ0 exp(2 t) + σ1 exp(−2 t) − t;
4
σ1 5
U (t) → u(x) = σ0 (2x + 3) + − log(2x + 3). (6.197)
(2x + 3) 4

X (B) : (x + 2)2 D 2 + 3(x + 2) D + 1 u(x) = (x + 2)2 log(x + 2)
(6.198)
Solution of Example X(B)

According to the schedule described earlier, set (x + 2) = exp(t) and (x + 2)D =
and transform (6.198) into the following:

[( − 1) + 3 + 1] U (t) = exp(2 t) log exp(t)

= 2 + 2 + 1 U (t) = exp(2 t) t. (6.199)
Now proceed with the usual routine. That is
E ch = k 2 + 2k + 1 = 0 ; k1,2 = − 1 ;
Scomp (t) = (σ0 + σ1 t) exp(−t) .
1 1
I pi (t) = 2 {exp(2 t) t} = exp(2 t) {t}
+ 2 + 1 ( + 2) + 2( + 2) + 1
2

1 exp(2 t) 6
= exp(2 t) {t} = 1− t
9 + 6 + 2 9 9

exp(2 t) 6
= I pi (t) = t−; U (t) = Scomp (t) + I pi (t)
9 9

exp(2 t) 6
= (σ0 + σ1 t) exp(−t) + t− .
9 9
σ0 + σ1 log(x + 2) (x + 2)2 6
U (t) → u(x) = + log(x + 2) − . (6.200)
(x + 2) 9 9

X (C) : (x − 1)2 D 2 − 1 u(x) = (x − 1) cos[log(x − 1)] log(x − 1).(6.201)
Solution of Example X(C)

According to the schedule described earlier, set (x − 1) = exp(t),
(x − 1)2 D 2 = ( − 1) and transform (6.201) into the following:
[( − 1) − 1] U (t) = exp(t) cos(t) t

2
= − − 1 U (t) = exp(t) cos(t) t. (6.202)
Again proceed with the usual routine.

√
1± 5
E ch = k 2 − k − 1 = 0 ; k1,2 = ;
# 2 $
√ √
t − 5 5
Scomp (t) = exp σ0 exp t + σ1 exp t ;
2 2 2
1 1 exp(i t) + exp(−i t)
I pi (t) = 2 {exp(t) cos(t) t} = {exp(t) t}.
−−1 −−1
2 2
1
= exp[t (1 + i)] {t}
( + 1 + i)2 − ( + 1 + i) − 1
1
+ exp[t (1 − i)] {t}.
( + 1 − i) − ( + 1 − i) − 1
2

1
=− exp[t (1 + i)] [(2 + i) − (1 − 2 i)] {t}
5

1
− exp[t (1 − i)] [(2 − i) − (1 + 2 i)] {t}.
5

1
=− exp[t (1 + i)] [(2 + i)t − (1 − 2 i)]
5

1
− exp[t (1 − i)] [(2 − i)t − (1 + 2 i)] .
5
exp(t)
= I pi (t) = [(2 + t) sin(t) + (1 − 2t) cos(t)] ;
5

# √ √ $
t − 5 5
= exp σ0 exp t + σ1 exp t
2 2 2
exp(t)
+ [(2 + t) sin(t) + (1 − 2t) cos(t)] ;
5 % √ √ &
1 − 5 5
U (t) → u(x) = (x − 1) 2 σ0 (x − 1) 2 + σ1 (x − 1) 2 +
1
(x − 1) {2 + log(x − 1)} sin[log(x − 1)] +
5
1
+ (x − 1) {1 − 2 log(x − 1)} cos[log(x − 1)] . (6.203)
5
6.9.4 Problems Group IX
Solve the following five problems by transforming each of the differential equations
by setting the variable (a x + b) = exp(a t). Then, work out their solution U (t).
Also write the corresponding solution u(x).

(x + 3)2 D 2 + 2(x + 3)D u = − log(x + 3) (x + 3) . (1)

(x + 3)2 D 2 + 2(x + 3)D + 3 u = − log(x + 3) (x + 3) . (2)

(2x − 1)2 D 2 − 2(2x − 1)D u = −2 log(2x − 1) sin[log(2x − 1)] (2x − 1) . (3)

(2x − 1)2 D 2 − 2(2x − 1)D + 4 u = −2 log(2x − 1) sin[log(2x − 1)] (2x − 1) . (4)
1
(3x + 2)2 D 2 − 3(3x + 2)D + u = sin[log(3x + 2)] (3x + 2) . (5)
4
6.10 Factorable Equations
Occasionally, there is a second- or higher-order linear (ODE) that, by good luck or

fortunate insight, can be factorized. Of course, the factorization has to be done in
such a manner that it keeps the differential equation operationally intact. For instance,
construct a second-order linear (ODE) with variable coefficients that is arranged in
advance to be operationally factorable into the following pair of first-order equations.
y1 = [D + M2 ]y(x), (6.204)
and
[x n D + M1 ]y1 = N , (6.205)
where for brevity M1 (x) has been represented as M1 , M2 (x) as M2 , N (x) as N , y1 (x)
as y1 . With the help of (6.204), which allows replacing y1 by [D + M2 ]y(x), (6.205)
can be represented as
[x n D + M1 ][D + M2 ]y(x) = N . (6.206)
The operator multiplication of the term on the left-hand side of (6.206) requires
some care. For instance, (6.206) leads to the following second-order (linODE) with
variable coefficients.
' (
x n D 2 + [M1 + x n M2 ] D + [x n (D M2 ) + M1 M2 ] y(x)
= N. (6.207)
In this chapter, for convenience, (6.204)–(6.207) are referred to as the ‘master

equations.’ Analyses of factorable second-order differential equations will make
extensive use of these master equations.
6.10 Factorable Equations 169
Unfortunately by comparing a second-order differential equation with its master

equation equivalent (6.207), one can glean only limited information. For instance,
while one can readily find what n and N are, determining M1 and M2 is less straight-
forward. Here, one knows only the sums [M1 + x n M2 ] and [x n (D M2 ) + M1 M2 ] .
But alas that does not a readily accessible solution of M1 and M2 make.
6.10.1 Examples Group XI
Check whether the following second-order equations are factorable. And if so, solve
them accordingly.

X I (1) : D 2 + (x −1 + 1) D + x −1 y(x) = x .
2
X I (2) : D + (x −1 + 3) D + 2(x −1 + 1) y = x −1 .

X I (3) : x D 2 − (2 x + 1)D + 2 y = x 3 exp(x) .
2
X I (4) : D + x −1 D − x −2 y = exp(x) .
2
X I (5) : D + 2x D + 2 y = log(x) .
2
X I (6) : D + exp(x) D + exp(x) y = 2 x exp(x) . (6.208)
Solution XI(1)
One-to-one comparison with (6.207) leads to the equalities
M1 + x n M2 = x −1 + 1 ; (6.209)
x n (D M2 ) + M1 M2 = x −1 . (6.210)
Because N = x and n = 0, M1 and M2 are determined by the relations
M1 + M2 = x −1 + 1. (6.211)
−1
(D M2 ) + M1 M2 = x . (6.212)
Elimination of M1 , (6.211) and (6.212), renders (6.212) into a Riccati equation—

compare (6.130)–(6.156)—
[D + (x −1 + 1)]M2 = [M2 ]2 + x −1 . (6.213)
and solving it would involve a second-order differential equation. Thus, one would
have made no headway at all ! Unless, of course, one were clever and could guess
M1 or M2 by inspection.
Another two possibilities for success are if

(A): M1 is vanishing
(B): or, M2 is a constant.
In order to solve (6.211) and (6.212), try the first possibility: meaning M1 = 0. If
this is true, then according to (6.211) and (6.212), M2 =
(x −1 + 1) and D M2 = x −1 . Which of course is not possible.
On the other hand, the second possibility, namely M2 is a constant = c, works.
Here, D M2 = 0 and (6.213) becomes
(x −1 + 1)c = c2 + x −1 . (6.214)
This is a quadratic with two solutions c = 1 and c = x −1 . Clearly the only acceptable
solution, where M2 = c is a constant, is c = 1. Knowing M2 = 1, (6.211), or (6.212),
gives M1 = x −1 .
Also because n = 0 and N = x, the central differential equations to be solved—
namely (6.205) ≡ (a) and (6.204)≡ (b)—can now be written as:
(a) : [D + x −1 ]y1 = x;
(b) : [D + 1] y(x) = y1 . (6.215)
Because both these equations are first-order (linODE), the procedures used in (4.2)–
(4.18) apply and lead readily to the relevant solution.
σ1 x2
(a) : y1 = + ;
x 3
exp (x) 1
(b) : y(x) = σ1 exp (−x) dx + σ2 exp (−x) + [x 2 − 2x + 2].
x 3
(6.216)
Solution XI(2)
Again work with the master equation (6.207) and the defining (6.204), (6.205),
and (6.206).
Here, n = 0, N = x −1 and
{2} : (M1 + M2 ) = (x −1 + 3) ; [D M2 + M1 M2 ] = 2(x −1 + 1).

(6.217)
Clearly, M1 = 0 does not work.

Next, try M2 = c. Then D M2 = 0. One of the two solutions of the resultant
quadratic in c is a constant equal to 2. This means c = M2 = 2 and therefore
M1 = (1 + x −1 ). Now that one knows all the relevant parameters, one can construct
the pair of first-order (linODE) that need solving. These are:
(a) : [D + (1 + x −1 )]y1 = x −1 ;
(b) : (D + 2)y(x) = y1 . (6.218)
In the usual manner, their solution is found to be the following:
exp(−x) 1
(a) : y1 = σ1 + ;
x x

exp(x)
(b) : y(x) = σ1 exp(−2x) dx + σ2 exp(−2x)
x

exp(2x)
+ exp(−2x) dx. (6.219)
x
Solution XI(3)
Next, solve differential equation {3}. Here, n = 1 and
{3} : (M1 + x M2 ) = −(2x + 1) ; [x(D M2 ) + M1 M2 ] = 2 . (6.220)
Therefore, M1 = 0 does not work.

Next, try M2 = c. Then, D M2 = 0 and (6.220) represent two relations: (M1 +
x c) = − (2x + 1) and c M1 = 2. Elimination of M1 leads to a quadratic in c
xc2 + (2x + 1)c + 2 = 0. (6.221)
Its two solutions are : c = −x −1 and c = −2. Clearly, the solution which is relevant
is when c is a constant. That is, c = −2. Knowing M2 = c = −2 one readily finds
M1 = −1. Knowing n, M1 , M2 , and the fact that N = x 3 exp(x), one can construct
(a) : [x D − 1]y1 = x 3 exp(x);

(b) : (D − 2)y(x) = y1 . (6.222)
(a) : y1 = σ1 x + (x 2 − x) exp(x) ;
σ1
(b) : y(x) = σ2 exp(2x) − (2x + 1)
2 4
− x + x + 1 exp(x) . (6.223)
Solution XI(4)
Here, n = 0, N = exp(x) and
{4} : (M1 + M2 ) = x −1 ; [D M2 + M1 M2 ] = − x −2 . (6.224)
As usual, first try M1 = 0. Therefore, M2 = x −1 . Then, D M2 = − x −2 and both

the equations work. Now, one can readily construct the relevant pair of first-order
(linODE). These are:
(a) : Dy1 = exp(x) ;

(b) : (D + x −1 )y(x) = y1 . (6.225)
Their solution is
(a) : y1 = σ1 + exp(x) ;
x σ2
(b) : y(x) = σ1 +
2 x

1
+ exp(x) − exp(x). (6.226)
x
Solution XI(5)
Here, n = 0, N = log(x) and
{5} : (M1 + M2 ) = 2x ; [D M2 + M1 M2 ] = 2. (6.227)
Set M1 = 0. Then, M2 = 2x and D M2 = 2. These two statements are consistent.

Now, one can readily construct the relevant pair of first-order (linODE). These are:
(a) : Dy1 = log(x);

(b) : (D + 2x)y(x) = y1 . (6.228)
Their solution is
(a) : y1 = σ1 − x + x log(x);

(b) : y(x) = σ1 exp(x 2 )dx + σ2 exp(−x 2 )

1 1 exp(x 2 )
+ [log(x) − 1] − exp(−x 2 ) dx. (6.229)
2 4 x
Solution XI(6)
Here, n = 0, N = 2x exp(x) and
{6} : (M1 + M2 ) = exp(x) ; [D M2 + M1 M2 ] = exp(x). (6.230)

Again, first try M1 = 0. Then, M2 = exp(x) and D M2 = exp(x). These two

statements are consistent. Therefore, the relevant pair of first-order linear (ODE) is
(a) : Dy1 = 2x exp(x) ;

(b) : D + exp(x) y(x) = y1 . (6.231)
Their solution is
(a) : y1 = σ1 + 2 exp(x) (x − 1);

(b) : y(x) = (σ1 − 2) exp {− exp(x)} exp {exp(x)}dx
+σ2 exp {− exp(x)} + 2x − 2 . (6.232)

∂u ∂v
u = 1 + 2y ; v = 1 + 3x. = 2 and = 3 . (6.233)
∂y x ∂x y
Because 2 = 3, this equation is inexact. Therefore, try using (6.99) to check as to

whether an integrating factor that depends only on the variable x is possible.
∂v
∂u
∂y
− ∂x y −dx dX (x)
x
.dx = = . (6.234)
v 1 + 3x X (x)
Good news! There exists an integrating factor X (x). One can find it by integrating
the above equation.
−1
log[X (x)] = log (1 + 3x) 3 + constant. (6.235)
Thus
σ0
X (x) = 1 . (6.236)
(1 + 3x) 3
Including the integrating factor, and ignoring the unnecessary multiplier σ0 , the
differential equation (6.104)-(1) now is
(1 + 3x)− 3 [(1 + 2y) dx + (1 + 3x) dy] = 0 .

1
(6.237)
Now determine whether this new differential equation is indeed exact. One has
u(x, y) = (1 + 2y)(1 + 3x)− 3 ; v(x, y) = (1 + 3x) 3 . T her e f or e,

1 2

∂u − 13 ∂v 2
(1 + 3x)− 3 3 . (6.238)
1
= 2(1 + 3x) ; =
∂y x ∂x u 3
Good! The new equation is exact. Accordingly, one can find its solution via the
procedure outlined in (6.88). That is

∂[ u(x, y) dx]
σ1 − σ2 = u(x, y) dx +v(x, y) dy − dy
∂y
x
− 13 2
= (1 + 3x) (1 + 2y)dx + (1 + 3x) 3 dy

∂[ u(x, y) dx]
− dy
∂y x
2
(1 + 2y)(1 + 3x) 3 2 2
= 2 + (1 + 3x) 3 y − (1 + 3x) 3 y
3 3
2
(1 + 2y)(1 + 3x) 3
= . (6.239)
2
−1 σ1 − σ2
y(x) = + 2 . (6.240)
2 (1 + 3x) 3
6.10.2 Problems Group VII

[Hint: Compare (6.81)–(6.128).]
(1) : 0 = dz = (2 + 3y) dx + (3 + 2x) dy.
(2) : 0 = dz = y 3 dx + 2x y 2 dy.

1
(3) : 0 = dz = x 2 + y dx + dy. (6.241)
3x
6.11 Additional Riccati Equations
A nonlinear differential equation of the form
dy
+ y 2 + f 1 (x) y + f 2 (x) = 0 (6.242)
dx
is sometime called a Riccati equation.
6.11 Additional Riccati Equations 175
6.11.1 Solution
Set
1 dz
= y . (6.243)
z dx
As a result,
dz
= zy. (6.244)
dx
Its differentiation gives
d2 z dy dz dy
2
= z +y =z + y2 z . (6.245)
dx dx dx dx
Now, multiply (6.242) by z
dy
z + y 2 + f 1 (x) y + f 2 (x) = 0 (6.246)
dx
dy
and replace z dx
according to (6.245). This gives
d2 z
− y 2 z + z y 2 + f 1 (x) z y + f 2 (x) z = 0 . (6.247)
dx 2
Remove −y 2 z + z y 2 and replace z y by dz

dx
. As a result, the Riccati equation (6.242)
is transformed into
d2 z dz
2
+ f 1 (x) + f 2 (x)z = 0 . (6.248)
dx dx
Notice that this differential equation, unlike the original Riccati equation, is linear.
But the removal of nonlinearity has resulted in increasing the order. And we now
have a second-order differential equation. But, because the new equation is linear,
there is a greater possibility of making progress. In particular, if f 1 (x) and f 2 (x)
should both turn out to be constants, (6.248) can be solved much like equations of
the type (3.1), (3.5), etc. And once z has been determined, y can be calculated by a
simple differentiation as is implicit in the form of (6.243).
6.11.2 Additional Examples Group VIII
First convert the Riccati equations given below into second-order linear equations
and then solve them.
(1) : dy
dx
+ y 2 + x y = 0.
(2) : dy
dx
+ y2 + x 2 y = 0.
(3) : dy
dx
+ y2 + y + 1 = 0.
(4) : dy
dx
+ y2 + 2 y + 2 = 0. (6.249)
Converting the Riccati equations into second-order linear ordinary differential equa-
tions is straightforward. All one needs to do is use (6.242) and (6.248). Accord-
ingly, (6.249) get transformed as follows:
(1) → (1 ) : d2 z
dx 2
+ x dxdz
=0
d2 z
(2) → (2 ) : dx 2
+ x 2 dx
dz
=0
(3) → (3 ) : d2 z
dx 2
+ dx dz
+z =0
2
(4) → (4 ) : d z
dx 2
+ 2 dx + 2 z
dz
=0 (6.250)
6.11.3 Solution
Equations (6.249)-(1) and (6.250)-(1’)

Introduce the notation p = dz
dx
and write (6.250)-(1’) as
dp
+ x p = 0. (6.251)
dx
Equation (6.251) is readily solved. [See, for instance, (4.2), (4.10) and (4.15) that
describe how such first-order linear ordinary differential equations are solved.] That
is
2
x
p = σ0 exp − . (6.252)
2
Because p = dz
dx
, its integration leads to z.
2
dz x
dx = z = p dx = σ0 exp − dx + σ1 . (6.253)
dx 2
And once z is known, y can be determined by using (6.243). That is
1 dz p σ0 exp − x2
y= = =
z dx z 2
σ0 exp − x2 dx + σ1
2
exp − x2
= 2
. (6.254)
exp − x2 dx + σ2
Equations (6.249)-(2) and (6.250)-(2’)

Follow the same procedure as in (6.251)–(6.254) and write (6.250)-(2’) as
dp
+ x2 p = 0 . (6.255)
dx
Again with the help of (4.2), (4.10), and (4.15) one gets
3
x
p = σ0 exp − . (6.256)
3
Because p = dz
dx
its integration leads to z.
3
x
dz = z = σ0 exp − dx + σ1 . (6.257)
3
And once z is known, y can be determined by using (6.243). Much like (6.254) one
can write y as
3
exp − x3
y= 3
. (6.258)
exp − x3 dx + σ2
Equations (6.249)-(3) and (6.250)-(3’)

Notice that (6.250)-(3’) is homogeneous linear ordinary differential equation with
constant coefficients. And according to (3.13), the complimentary solution of a
second-order homogeneous linear ordinary differential equation with constant coef-
ficients can be written as
z ≡ Scomp (x) = σ1 exp (k1 x) + σ2 exp (k2 x). (6.259)

Here, k = k1 and k = k2 are solutions of the characteristic equation E ch for (6.250)-

2
dz
(3’). The procedure for finding E ch is to replace z by 1 , dx by k , and ddz z2 by k 2 .
Thus, the complementary solution for (6.250)-(3’) is found by solving the following
E ch .
k2 + k + 1 = 0 (6.260)
which leads to
√
−1 + i 3
k = k1 = r + im = ,
2 √
−1 − i 3
k = k2 = r − im = . (6.261)
2
Therefore, using (6.259)–(6.261), the result for z, according to (3.21) and (3.22), can
be written as
z = σ1 exp (r − i m) x + σ2 exp (r + i m)x

= exp (r x) [σ3 sin(m x) + σ4 cos(m x)]
√ √
x 3x 3x
= exp − σ3 sin + σ4 cos . (6.262)
2 2 2
Or, alternatively, as
√
x 3
z = σ0 exp − cos σ5 − x . (6.263)
2 2
Note, one went from (6.262)–(6.263) by introducing two new arbitrary constants σ0
and σ5 such that σ3 = σ0 sin(σ5 ) and σ4 = σ0 cos(σ5 ). This is perfectly alright as
long as σ02 = σ32 + σ42 . And because all sigmas are arbitrary constants, this equality
is trivially achieved.
Given z—see, (6.263)—the solution y is readily found.
√ √
1 dz 1 3 3
y = = − + tan σ5 − x . (6.264)
z dx 2 2 2
Equations (6.249)-(4) and (6.250)-(4’)

Again follow a similar procedure to that used for (6.259)–(6.264) and write the E ch
for (6.250)-(4’) as
k 2 + 2k + 2 = 0 (6.265)
which leads to
k = k1 = r + im = − 1 + i,
k = k2 = r − im = − 1 − i. (6.266)
Therefore,
z = exp (−x) [σ3 sin x + σ4 cos x]

≡ σ0 exp (−x) cos (σ5 − x) . (6.267)
And the solution y is
1 dz
y = = − 1 + tan (σ5 − x). (6.268)
z dx
6.11.4 Additional Problems Group VIII
Solve the following two problems that involve Riccati equations.

[Hint: Read (6.249)–(6.268).]
dy
(1) : + y 2 + 2 y + 1 = 0.
dx
dy
(2) : + y 2 + y + 2 = 0.
dx
6.12 Additional Euler Equations
Among Euler’s multifarious contributions to mathematics, in particular to the theory

of differential equations, is his work related to linear ordinary differential equations
with constant coefficients. H e7. showed how linear ordinary differential equations
with variable coefficients of the form, cn x n , given below
ν
"
cn x n D n u = B(x), (6.269)
n=0
could be transformed into linear ordinary differential equations with constant coeffi-
cients. The latter are much easier to solve as was demonstrated in detail in Chaps. 3
and 4.
The Euler procedure consists in arranging an appropriate change of the indepen-
dent variable that transforms the operator D into a new operator . One proceeds as
follows. Set
x = exp(t)
and differentiate with respect to x.
dt
1 = exp(t) . . (6.270)
dx

Multiply both sides of (6.270) from the right by d Ud(x,t)
t
.
d U (x, t) dt dU (x, t)
1· = exp(t) . .
dt dx dt
d U (x, t)
= exp(t) ·
dx
d U (x, t) d U (x, t)
= =x . (6.271)
dt dx
Because U (x, t) is arbitrary, (6.271) implies x d

dx
= d
dt
: meaning
x D = . (6.272)
Next, let us look at x D (x D).
x D (x D) = x (D x) D + x 2 D 2
= () = x D + x 2 D 2 = + x 2 D 2 . (6.273)
Therefore
x 2 D 2 ≡ ( − 1) . (6.274)
Similarly

x D x 2 D 2 = [ ( − 1)] . (6.275)
One also has

x D x 2 D2 = x 2 x D2 + x 3 D3 = 2 x 2 D2 + x 3 D3
= 2 ( − 1) + x 3 D 3 . (6.276)
The left-hand sides of (6.275) and (6.276) are equal. The equality of their right-hand
sides yields
6.12 Additional Euler Equations 181
2 ( − 1) + x 3 D 3 = [ ( − 1)] , (6.277)
or
x 3 D 3 ≡ ( − 1)( − 2). (6.278)
In order to proceed to the next order, multiply x 3 D 3 from the left by x D and
use (6.278) to appropriately replace one of the expressions equal to 3 x 3 D 3 .

x D x 3 D3 = x 3 x 2 D3 + x 4 D4 = 3 x 3 D3 + x 4 D4
= 3 ( − 1) ( − 2) + x 4 D 4 . (6.279)
Multiply the left-hand side of (6.278) by x D and the right-hand side by the same
amount, that is . One gets

x D x 3 D 3 = [( − 1)( − 2)] . (6.280)
Now that the left-hand sides of (6.279) and (6.280) are the same, one can claim
equality of their right-hand sides. That is
3 ( − 1) ( − 2) + x 4 D 4 = [( − 1)( − 2)] ,
which leads to the result
x 4 D 4 ≡ ( − 1)( − 2)( − 3). (6.281)
Exercise
Clearly, (6.272), (6.274), (6.278), and (6.281) show a pattern. Therefore, by induc-
tion, one assumes
x n D n = ( − 1)...( − n + 1). (6.282)
6.12.1 Additional Examples IX: Euler Equation
Transform each of the four differential equations I X − (A) → I X − (D) given

below by using the transformation x = exp(t), and renaming the function u(x) as
U (t). Finally, represent the result in terms of x. Then, find u(x) .

I X − (A) : x 2 D 2 + x D + 1 u(x) = a x 2 + b. (6.283)
6.12.2 Solution Additional Examples IX-(A)

Use (6.270), (6.272) and (6.274) and transform (6.283) into

[( − 1) + + 1] U (t) = 2 + 1 U (t) = a exp(2 t) + b. (6.284)
Then, solve for U (t) by following the familiar routine.
E ch = k 2 + 1 = 0 ; k1,2 = ±i;
Scomp (t) = σ0 exp(−i t) + σi exp(−i t) = σ2 sin(t) + σ3 cos(t);
1 1 b
I pi (t) = 2 {a exp(2 t) + b} = a exp(2 t) 2 + ;
+1 2 +1 1
a
U (t) = Scomp (t) + I pi (t) = σ2 sin(t) + σ3 cos(t) + exp(2 t) + b.
5
Finally, transform U (t) back into u(x). In this fashion one gets:
a 2
U (t) → u(x) = σ2 sin[log(x)] + σ3 cos[log(x)] + x + b. (6.285)
5
6.12.3 Solution Additional Examples IX-(B)

I X − (B) : x 2 D 2 − 2 x D + 4 u = a log(x). (6.286)
As above use (6.270), (6.272), and (6.274) and transform (6.286) into the following
differential equation.
2
− 3 + 4 U (t) = a t. (6.287)
Its solution is found in the usual manner.

√
3 7
E ch = k 2 − 3 k + 4 = 0 ; k1,2 = ± i ;
2 2
√ √
3t 7 3t 7
Scomp (t) = σ1 exp sin t + σ2 exp cos t ;
2 2 2 2

1 1 3 a 3a
I pi (t) = 2 (a t) = + (a t) = t + ;
− 3 + 4 4 16 4 16
U (t) = Scomp (t) + I pi (t) ;
√ √
7 3 7 3
U (t) → u(x) = σ1 sin log(x) x + σ2 cos
2 log(x) x 2
2 2
a 3a
+ log(x) + . (6.288)
4 16
Solution Example IX-(C)

I X − (C) : x 2 D 2 + 2x D − 4 u = a x 2 log(x) + b . (6.289)
Again use (6.270), (6.272), and (6.274) and transform (6.289) into
2
+ − 4 U (t) = a exp(2 t) t + b, (6.290)
Its solution is as follows:

√
1 17
E ch = k 2 + k − 4 = 0
; k1,2 = − ± ;
2 2
√ √
t 17 t 17
Scomp (t) = σ1 exp − + t + σ2 exp − − t ;
2 2 2 2
1 1
I pi (t) = 2 {a exp(2 t) t + b} = a exp(2t) t
+−4 { + 2) + ( + 2) − 4}
2

b 1 b 1 5 b
+ = a exp(2t) t − = a exp(2t) − t− ;
−4 2 + 5 + 2 4 2 4 4
1 5 b
I pi (t) = a exp(2t) t − − ;
2 4 4
U (t) = Scomp (t) + I pi (t);
√ √
1 17 1 17
U (t) → u(x) = σ1 x − 2 + 2 + σ2 x − 2 − 2
1 5
+ a x2 log(x) − . (6.291)
2 4
-
Solution Example IX-(D)

I X − (D) : x 2 D 2 − 4x D + 2 u = 2 x cos log(x) . (6.292)
And finally, as before, use (6.270), (6.272), and (6.274) and transform (6.292) into
2
− 5 + 2 U (t) = 2 exp(t) cos(t). (6.293)
This leads to
√
5 17
E ch = k 2 − 5 k + 2 ; k1,2 = ± ;
2 2
√ √
5 17 5 17
Scomp (t) = σ1 exp t+ t + σ2 exp t− t ;
2 2 2 2
1 1
I pi (t) = 2 {2 a exp(t) cos(t)} = exp[t (1 + i)]
− 5 + 2 2 − 5 + 2
1
+ 2 exp[t (1 − i)];
− 5 + 2
1 1
= exp[t (1 + i)] + exp[t (1 − i)]
−3 − 3i −3 + 3i
exp(t) 1
=− (1 − i) exp(i t) + (1 + i) exp(−i t) = I pi (t) = − exp(t) [sin(t) + cos(t)] .
6 3

√ √
5 17 5 17
U (t) → u(x) = σ1 x 2 + 2 + σ2 x 2 − 2
x
− sin{log(x)} + cos{log(x)} . (6.294)
3
6.12.4 Additional Extended Euler Equations
A simple extension of Euler equation is given below in (6.295).

ν
"
cn (a x + b)n D n u = B(x), (6.295)
n=0
Any such (linODE) with variable coefficients of the form cn (a x + b)n can readily
be transformed into a (linODECC)—meaning into equations of the type that we have
been studying previously. The relevant transformation is carried out by a change of
the variable: from (a x + b) to exp(a t) where both a and b are constants. That is
(a x + b) = exp(a t). (6.296)
Differentiate (6.296) with respect to x , and divide both sides by the constant a.
dt
1 = exp(a t) .
dx
dt
= (a x + b) . . (6.297)
dx

Multiply both sides of (6.297) from the right by dUdt(x,t) .
dU (x, t) dt dU (x, t)
= (a x + b) .
dt dx dt
dU (x, t)
= (a x + b) . . (6.298)
dx
Because U (x, t) is arbitrary, this implies d

dt
= (a x + b) dxd : meaning
= (a x + b) D. (6.299)
Next, let us look at (a x + b) D {(a x + b) D} or in other words at {}.
(a x + b) D {(a x + b) D} = (a x + b)[a D + (a x + b) D 2 ]
= {} = a + (a x + b)2 D 2 . (6.300)
Transferring a across to the left side gives
(a x + b)2 D 2 = ( − a) . (6.301)
Similarly

(a x + b) D (a x + b)2 D 2 = (a x + b) 2a 2 x + 2ab D 2 + (a x + b)[(a x + b)2 D 3 ]
= 2a(ax + b)2 D 2 + (ax + b)3 D 3

= 2 a ( − a) + (a x + b)3 D 3 . (6.302)
By using (6.301) and the equivalence (a x + b) D = on the left-hand side of

(6.302), one gets
[ ( − a)] = 2 a ( − a) + (a x + b)3 D 3 , (6.303)
which leads to
(a x + b)3 D 3 = ( − a)( − 2 a). (6.304)

Again, starting with (a x + b) D (a x + b)3 D 3 , one can readily show that
(a x + b)4 D 4 = ( − a)( − 2 a)( − 3 a). (6.305)
Also, in view of (6.299), (6.301), (6.304), (6.305), or equivalently (6.306) given

below, one can transform a general equation of the type (6.295) into one involving
’s and functions of the variable t. One must not forget, however, to use (ax + b) =
exp(at) for transforming out of the variable B(x) into an appropriate function of the
variable t. This will become clear in examples (X I ) given below.
Exercise
Clearly, (6.299), (6.301), (6.304), and (6.305) show a pattern. Therefore, by induction
(a x + b)n D n = { − a}...{ − (n − 1)a}. (6.306)
6.12.5 Additional Examples Group X: Extended Euler

Equations
Transform each of the following three differential equations by setting the variable
(a x + b) = exp(a t), and then, find their solution U (t). Also write the correspond-
ing solution u(x).

X (A) : (2x + 3)2 D 2 + 2(2x + 3) D − 4 u(x) = 5 log[2x + 3]. (6.307)
Solution of Example X(A)

Set (2x + 3) = exp(2 t) and (2x + 3)D = , use (6.299) and (6.301), and trans-
form (6.307) into the following:

[( − 2) + 2 − 4] U (t) = 5 log exp(2t)

= 2 − 4 U (t) = 10 t. (6.308)
Now, proceed with the usual routine.
E ch = k 2 − 4 = 0 ; k1,2 = ±2;
Scomp (t) = σ0 exp(2 t) + σ1 exp(−2 t) ;
1 1 2
I pi (t) = 2 {10 t} = − 1 + {10 t}
−4 4 4
10
=− t ;
4
10
U (t) = Scomp (t) + I pi (t) = σ0 exp(2 t) + σ1 exp(−2 t) − t;
4
σ1 5
U (t) → u(x) = σ0 (2x + 3) + − log(2x + 3). (6.309)
(2x + 3) 4

X (B) : (x + 2)2 D 2 + 3(x + 2) D + 1 u(x) = (x + 2)2 log(x + 2)
(6.310)
Solution of Example X(B)

According to the schedule described earlier, set (x + 2) = exp(t) and (x + 2)D =
and transform (6.310) into the following:

[( − 1) + 3 + 1] U (t) = exp(2 t) log exp(t)

= 2 + 2 + 1 U (t) = exp(2 t) t. (6.311)
Now proceed with the usual routine. That is
E ch = k 2 + 2k + 1 = 0 ; k1,2 = −1;
Scomp (t) = (σ0 + σ1 t) exp(−t) .
1 1
I pi (t) = 2 {exp(2 t) t} = exp(2 t) {t}
+ 2 + 1 ( + 2)2 + 2( + 2) + 1

1 exp(2 t) 6
= exp(2 t) {t} = 1− t
9 + 6 + 2 9 9

exp(2 t) 6
= I pi (t) = t− ; U (t) = Scomp (t) + I pi (t)
9 9

exp(2 t) 6
= (σ0 + σ1 t) exp(−t) + t− .
9 9
σ0 + σ1 log(x + 2) (x + 2)2 6
U (t) → u(x) = + log(x + 2) − . (6.312)
(x + 2) 9 9

X (C) : (x − 1)2 D 2 − 1 u(x) = (x − 1) cos[log(x − 1)] log(x − 1).(6.313)
Solution of Example X(C)

According to the schedule described earlier, set (x − 1) = exp(t) ,
(x − 1)2 D 2 = ( − 1) and transform (6.313) into the following:
[( − 1) − 1] U (t) = exp(t) cos(t) t

2
= − − 1 U (t) = exp(t) cos(t) t. (6.314)
Again proceed with the usual routine.

√
1± 5
E ch = k 2 − k − 1 = 0; k1,2 = ;
# 2
√ $
√
t − 5 5
Scomp (t) = exp σ0 exp t + σ1 exp t ;
2 2 2
1 1 exp(i t) + exp(−i t)
I pi (t) = 2 {exp(t) cos(t) t} = {exp(t) t}.
−−1 −−1
2 2
1
= exp[t (1 + i)] {t}
( + 1 + i)2 − ( + 1 + i) − 1
1
+ exp[t (1 − i)] {t}.
( + 1 − i) − ( + 1 − i) − 1
2

1
=− exp[t (1 + i)] [(2 + i) − (1 − 2 i)] {t}
5

1
− exp[t (1 − i)] [(2 − i) − (1 + 2 i)] {t}.
5

1
=− exp[t (1 + i)] [(2 + i)t − (1 − 2 i)]
5

1
− exp[t (1 − i)] [(2 − i)t − (1 + 2 i)] .
5
exp(t)
= I pi (t) = [(2 + t) sin(t) + (1 − 2t) cos(t)] ;
5

# √ √ $
t − 5 5
= exp σ0 exp t + σ1 exp t
2 2 2
exp(t)
+ [(2 + t) sin(t) + (1 − 2t) cos(t)] ;
5 % √ √ &
1 − 5 5
U (t) → u(x) = (x − 1) 2 σ0 (x − 1) 2 + σ1 (x − 1) 2
1
+ (x − 1) {2 + log(x − 1)} sin[log(x − 1)] +
5
1
+ (x − 1) {1 − 2 log(x − 1)} cos[log(x − 1)]. (6.315)
5
6.12.6 Additional Problems Group IX
Solve the following five problems by transforming each of the differential equations
by setting the variable (a x + b) = exp(a t). Then, work out their solution U (t).
Also write the corresponding solution u(x).

(x + 3)2 D 2 + 2(x + 3)D u = − log(x + 3) (x + 3) . (1)

(x + 3)2 D 2 + 2(x + 3)D + 3 u = − log(x + 3) (x + 3) . (2)

(2x − 1)2 D 2 − 2(2x − 1)D u = −2 log(2x − 1) sin[log(2x − 1)] (2x − 1) . (3)

(2x − 1)2 D 2 − 2(2x − 1)D + 4 u = −2 log(2x − 1) sin[log(2x − 1)] (2x − 1) . (4)
1
(3x + 2)2 D 2 − 3(3x + 2)D + u = sin[log(3x + 2)] (3x + 2) . (5)
4
6.13 Additional Factorable Equations
Occasionally there is a second- or higher-order (linODE) that, by good luck or fortu-

nate insight, can be factorized. Of course, the factorization has to be done in such a
manner that it keeps the differential equation operationally intact. For instance, con-
struct a second-order (linODE) with variable coefficients that is arranged in advance
to be operationally factorable into the following pair of first-order equations.
y1 = [D + M2 ]y(x), (6.316)
and
[x n D + M1 ]y1 = N , (6.317)
where for brevity M1 (x) has been represented as M1 , M2 (x) as M2 , N (x) as N , y1 (x)
as y1 . With the help of (6.316), which allows replacing y1 by [D + M2 ]y(x), (6.317)
can be represented as
[x n D + M1 ][D + M2 ]y(x) = N . (6.318)
The operator multiplication of the term on the left-hand side of (6.318) requires
some care. For instance, (6.318) leads to the following second-order (linODE) with
variable coefficients.
' (
x n D 2 + [M1 + x n M2 ] D + [x n (D M2 ) + M1 M2 ] y(x)
= N. (6.319)
In this chapter, for convenience, (6.316)–(6.319) are referred to as the ‘master

equations.’ Analyses of factorable second-order differential equations will make
extensive use of these master equations.
Unfortunately by comparing a second-order differential equation with its master
equation equivalent (6.319), one can glean only limited information. For instance,
while one can readily find what n and N are, determining M1 and M2 is less straight-
forward. Here, one knows only the sums [M1 + x n M2 ] and [x n (D M2 ) + M1 M2 ].
But alas that does not a readily accessible solution of M1 and M2 make.
6.13.1 Examples Group XI: Additional Factorable Equations
Check whether the following second-order equations are factorable. And if so, solve
them accordingly.

X I (1) : D 2 + (x −1 + 1) D + x −1 y(x) = x.
2
X I (2) : D + (x −1 + 3) D + 2(x −1 + 1) y = x −1 .

X I (3) : x D 2 − (2 x + 1)D + 2 y = x 3 exp(x).
2
X I (4) : D + x −1 D − x −2 y = exp(x).
2
X I (5) : D + 2x D + 2 y = log(x).
2
X I (6) : D + exp(x) D + exp(x) y = 2 x exp(x). (6.320)
Solution: Example XI(1)

One-to-one comparison with (6.319) leads to the equalities
M1 + x n M2 = x −1 + 1 ; (6.321)
−1
x (D M2 ) + M1 M2 = x
n
. (6.322)
Because N = x and n = 0, M1 and M2 are determined by the relations
M1 + M2 = x −1 + 1. (6.323)
−1
(D M2 ) + M1 M2 = x . (6.324)
Elimination of M1 , from (6.323) and (6.324), renders (6.324) into a Riccati

equation—compare (6.242)–(6.268)—
[D + (x −1 + 1)]M2 = [M2 ]2 + x −1 . (6.325)
and solving it would involve a second-order differential equation. Thus, one would
have made no headway at all ! Unless, of course, one were clever and could guess
M1 or M2 by inspection.
Another two possibilities for success are if
(A): M1 is vanishing
(B): or, M2 is a constant.
In order to solve (6.323) and (6.324), try the first possibility: meaning M1 = 0. If
this is true, then according to (6.323) and (6.324), M2 = (x −1 + 1) and D M2 = x −1 .
6.13 Additional Factorable Equations 191
Which of course is not possible.

On the other hand, the second possibility, namely M2 is a constant = c , works. Here,
D M2 = 0 and (6.325) becomes
(x −1 + 1)c = c2 + x −1 . (6.326)
This is a quadratic with two solutions c = 1 and c = x −1 . Clearly, the only acceptable
solution, where M2 = c is a constant, is c = 1. Knowing M2 = 1, (6.323), or (6.324),
gives M1 = x −1 .
Also because n = 0 and N = x, the central differential equations to be solved—

namely (6.317)≡ (a) and (6.316)≡ (b)—can now be written as:
(a) : [D + x −1 ]y1 = x;
(b) : [D + 1] y(x) = y1 . (6.327)
Because both these equations are first-order (linODE), the procedures used in (4.2)–
(4.18) apply and lead readily to the relevant solution.
σ1 x2
(a) : y1 = + ;
x 3
exp (x) 1
(b) : y(x) = σ1 exp (−x) d x + σ2 exp (−x) + [x 2 − 2x + 2].
x 3
(6.328)

Again work with the master equation (6.319) and the defining (6.316),
(6.317), and (6.318).
Here, n = 0, N = x −1 and
{2} : (M1 + M2 ) = (x −1 + 3) ; [D M2 + M1 M2 ] = 2(x −1 + 1). (6.329)
Clearly, M1 = 0 does not work.

Next, try M2 = c. Then D M2 = 0. One of the two solutions of the resultant
quadratic in c is a constant equal to 2. This means c = M2 = 2 and therefore
M1 = (1 + x −1 ). Now that one knows all the relevant parameters, one can construct
(a) : [D + (1 + x −1 )]y1 = x −1 ;
(b) : (D + 2)y(x) = y1 . (6.330)
exp(−x) 1
(a) : y1 = σ1 + ;
x x

exp(x)
(b) : y(x) = σ1 exp(−2x) dx + σ2 exp(−2x)
x

exp(2x)
+ exp(−2x) dx. (6.331)
x

Next, solve differential equation {3}. Here, n = 1 and
{3} : (M1 + x M2 ) = −(2x + 1) ; [x(D M2 ) + M1 M2 ] = 2 . (6.332)
Therefore, M1 = 0 does not work.

Next, try M2 = c. Then D M2 = 0 and (6.332) represent two relations: (M1 +
x c) = − (2x + 1) and c M1 = 2. Elimination of M1 leads to a quadratic in c
xc2 + (2x + 1)c + 2 = 0 . (6.333)
Its two solutions are : c = −x −1 and c = −2. Clearly, the solution which is relevant
is when c is a constant. That is, c = −2. Knowing M2 = c = −2, one readily finds
M1 = −1. Knowing n, M1 , M2 , and the fact that N = x 3 exp(x), one can construct
(a) : [x D − 1]y1 = x 3 exp(x) ;

(b) : (D − 2)y(x) = y1 . (6.334)
(a) : y1 = σ1 x + (x 2 − x) exp(x) ;
σ1
(b) : y(x) = σ2 exp(2x) − (2x + 1)
2 4
− x + x + 1 exp(x) . (6.335)

Here, n = 0 , N = exp(x) and
{4} : (M1 + M2 ) = x −1 ; [D M2 + M1 M2 ] = − x −2 . (6.336)

6.13 Additional Factorable Equations 193
As usual, first try M1 = 0. Therefore, M2 = x −1 . Then, D M2 = − x −2 and both

the equations work. Now, one can readily construct the relevant pair of first-order
(linODE). These are:
(a) : Dy1 = exp(x) ;

(b) : (D + x −1 )y(x) = y1 . (6.337)
Their solution is
(a) : y1 = σ1 + exp(x) ;
x σ2
(b) : y(x) = σ1 +
x
2
1
+ exp(x) − exp(x). (6.338)
x

Here, n = 0, N = log(x) and
{5} : (M1 + M2 ) = 2x ; [D M2 + M1 M2 ] = 2. (6.339)
Set M1 = 0. Then, M2 = 2x and D M2 = 2. These two statements are consistent.

Now, one can readily construct the relevant pair of first-order (linODE). These are:
(a) : Dy1 = log(x) ;

(b) : (D + 2x)y(x) = y1 . (6.340)
Their solution is
(a) : y1 = σ1 − x + x log(x);

(b) : y(x) = σ1 exp(x 2 )dx + σ2 exp(−x 2 )

1 1 exp(x 2 )
+ [log(x) − 1] − exp(−x 2 ) dx.
2 4 x
(6.341)

Here, n = 0 , N = 2x exp(x), and
{6} : (M1 + M2 ) = exp(x) ; [D M2 + M1 M2 ] = exp(x). (6.342)

Again, first try M1 = 0. Then, M2 = exp(x) and D M2 = exp(x). These two

statements are consistent. Therefore, the relevant pair of first-order (linODE) is
(a) : Dy1 = 2x exp(x) ;

(b) : D + exp(x) y(x) = y1 . (6.343)
Their solution is
(a) : y1 = σ1 + 2 exp(x) (x − 1) ;

(b) : y(x) = (σ1 − 2) exp {− exp(x)} exp {exp(x)}dx
+ σ2 exp {− exp(x)} + 2x − 2 . (6.344)
Chapter 7
Special Situations
Unlike equations of the first-order and first-degree linear (ODE)’s that are second and
higher order, and have variable coefficients, are often difficult to solve. Fortunately,
there are equations of special types that are easier to handle. Some of these were
treated in the preceding chapter. Different from special types, but somewhat in the
same vein, are equations that represent ‘Special Situations’. With such situations
in place, satisfactory solution is often possible. For instance, a given differential
equation may be integrable. Similarly, equations that have both the independent
as well as the dependent variables missing in any explicit form: And those that
explicitly contain only the independent variable, or only the dependent variable, are
easily handled.
An interesting new situation called order reduction comes into play if one of the
n non-trivial solution of an nth order homogeneous linear (ODE) is already known.
Then, the given equation can be reduced to one of (n − 1)th order.
7.1 Ordinary Differential Equations
Both homogeneous—[See (3.1)–(3.56)]—and inhomogeneous—[See (3.57)–(3.91)]

—first- and second-order linear ordinary differential equations that have constant
coefficients were discussed in detail in Chap. 3. Indeed, that discussion would also
apply to similar equations of order higher than the second. Also studied, there were
simultaneous linear ordinary differential equations—[See (3.92)–(3.143)]. But owing
to the complexity of this latter study, simultaneous equations involving only two or
three dependent variables were treated.
Chapter 4 dealt with equations of first order and first degree with variable coeffi-
cients. Included there was a discussion of the special case of the Bernouilli equations

https://doi.org/10.1007/978-3-319-76406-1_7
196 7 Special Situations
[See (4.19)–(4.77)]. These equations, it turns out, can be transformed into standard
linear (ODE)) of first order and first degree.
The Second Order
Unlike equations of the first order and first degree, as a rule linear (ODE)’s that
are second and higher order and have variable coefficients are difficult to solve
in terms of known elementary functions. Fortunately, there are exceptions to this
gloomy rule. But, much like the Ber nouilli 1. equations, mostly the exceptions are
equations of special types. Some examples of these are special equations that were
described in the preceding chapter. Included there were the Clairaut 2. equations—
[Compare (6.2)–(6.13)]—the Lagrange3. equation—[Compare (6.19)–(6.31)]—the

Separable Equations—[Compare (6.32)–(6.35)] and the dy dx
= xy equations—
[Compare (6.36)–(6.73)]. In addition, there are the so-called Exact- [Compare (6.74)–
(6.91)] and Inexact-Equations—[Compare (6.92)–(6.241)]—Riccati 4. equations—
[Compare (6.242)–(6.268)]—Euler 7. equations—[Compare (6.269)–(6.315)] , and
the Factorable Equations—[Compare (6.316)–(6.344)].
Described below are a few equations that represent special situations. Such situa-
tions are slightly different from the special types treated in the preceding chapter. And
with these situations in place, satisfactory solution of relevant differential equations
is often possible.
Simple Cases—see (7.1)–(7.27)—are studied first. These cases are either readily
integrated, or they have the independent and/or the dependent variables missing in
explicit form.
In (7.29)–(7.47), one works with an interesting new situation that is referred to
as order reduction. Generally, it is not very likely that one of the non-trivial solution
to a given differential equation is known. However, if it should happen, the given
equation can be reduced to one which is of order one lower than the original. This
is particularly helpful when dealing with a first-degree, second-order homogeneous
linear (ODE). Because with prior knowledge of one non-trivial solution, the equation
evolves into one of the first order. And such first-order differential equations are often
more manageable than the original second-order equations.
Should a differential equation with variable coefficients also be inhomogeneous,
finding its complete solution would require knowledge of both its complementary
solution and the particular integral. While the method of undetermined coefficients—
see (3.60)–(3.91)—worked well for determining the I pi for equations with constant
coefficients, differential equations with variable coefficients, in contrast, are best
treated by a procedure termed variation of parameters . This latter procedure is
described and extensively illustrated by relevant application—see (7.48) → (7.62).
Finally, two different varieties of second-order differential equations are con-
sidered where a known special situation obtains. For known special situations, the
second-order differential equation can often be solved by following particular rou-
tines which are special to that case. These routines are described in detail in (7.63)–
(7.124).
7.2 Simple Cases 197
7.2 Simple Cases
Equation (1)
Consider
n
D2 y = (x) , (7.1)
or, equivalently, the second-order, first-degree ordinary differential equation

2
(1) : D y = f (x) . (7.2)
Here, as usual, D = dx d
. Clearly, differential equations of the form (7.2) can directly
be integrated. First and second integration yields

[D 2 y]dx = Dy = f (x)dx + σ1 ,

[Dy]dx = y = dx f (x)dx + σ1 x + σ2 . (7.3)
As an exercise, in (7.2), try
f (x) = x0 − x . (7.4)
Then, (7.3) leads to the result

x0 x 2 x3
y= − + σ1 x + σ2 . (7.5)
2 6
Equation (2)
Next, treat equations of the form
(2) : D 2 y(x) = f (D y(x)) . (7.6)
Use the notation D = d

dx
, q(x) = dy(x)
dx
= D y(x), and rewrite (7.6) as
dq(x)
= f [q(x)]
dx
and integrate.

dq(x)
= x + constant . (7.7)
f [q(x)]
The left-hand side of (7.7) is some function F[q(x)] that in principle determines q(x).
Once q(x), = dy(x)
dx
, is known, its integration would lead to the desired solution y(x).
Equation (3)
2
D2 y = 1 + (Dy)2 . (7.8)
A more suitable form, convenient for calculation, is the square root of (7.8).
d q(x)
(3) : D 2 y(x) = Dq(x) =
dx
= ± 1 + [q(x)]2 . (7.9)
Or, equivalently,
dq(x)
= ± dx . (7.10)
1 + [q(x)]2
Integration leads to
sinh−1 [q(x)] = ±x + constant . (7.11)
Inverting (7.11) yields
dy(x)
q(x) = = sinh [x + σ1 ] . (7.12)
dx
And final integration with respect to x leads to
y(x) = cosh (x + σ1 ) + σ2 . (7.13)
Solve the following (ODE).

2
D2 y = 1 + (Dy) .
7.2.2 Equations (a) and (b)
Consider second-order, first-degree linear (ODE) of the form
D 2 y(x) + (x) Dy(x) = 0 .

This equation is best re-expressed as a first-order, first-degree linear (ODE) in terms

of q(x)—meaning, dy(x)dx
—as the new dependent variable.
Dq(x) + (x) q(x) = 0 , (7.14)
Once solved, q(x) can be integrated with respect to x to find y(x). Below, we solve
the following first-degree linear (ODE)’s.
(a) : D 2 y + (x + x 2 ) Dy = 0 ;
(b) : D 2 y + exp(x) Dy = 0 . (7.15)
In the following, we define M(x), N (x), and W (x) and work out the solution.
7.2.3 Solution
(a) : Dq(x) + (x + x 2 ) q(x) = 0 .

M(x) ≡ x + x 2 ; N (x) = 0 ;

x2 x3
W (x) ≡ exp M(x) dx = exp + ;
2 3

1
q(x) ≡ W (x) N (x) dx + σ1
W (x)
x2 x3
= σ1 exp − − .
2 3
Integrating q(x) with respect to x leads to the desired solution.

x2 x3
y(x) = q(x) dx = σ1 exp − − dx + σ2 . (7.16)
2 3
(b) : Dq(x) + exp(x) q(x) = 0 .
M(x) ≡ exp(x) ; N (x) = 0 ;

W (x) ≡ exp M(x) dx = exp {exp(x)} ;

1
q(x) ≡ W (x) N (x) dx + σ1
W (x)
= σ1 exp {− exp(x)} .
Therefore, the solution y(x) is

y(x) = q(x)dx = σ1 exp {− exp(x)}dx + σ2 . (7.17)
Clearly, this procedure would work just as well for other choicesof the function (x).
To that end, it is helpful to set z = − exp(x). Then, represent exp {− exp(x)}dx

as exp(z)
z
dz = σ3 Ei[z] = σ3 Ei[-exp(x)].
7.2.4 Equations (c) and (d)
Solve the following second-order, first-degree linear (ODE).
1
(c) : D 2 y(x) + Dy(x) = 0 .
y(x)
(d) : D 2 y(x) + 2 y(x) . Dy(x) = 0 . (7.18)
Solution (c)
Because
dq(x) dq(x) dy(x) dq(x)
D 2 y(x) = Dq(x) = = . = . q(x) , (7.19)
dx dy(x) dx dy(x)
and
dq(x) 1
(c) : . q(x) + . Dy(x)
dy(x) y(x)
dq(x) 1
= . q(x) + . q(x) = 0 . (7.20)
dy(x) y(x)
Therefore,
dq(x) 1
(c) : =− . (7.21)
dy(x) y(x)
Integrate the above with respect to y(x) .

dq(x) dy(x)
. dy(x) = dq(x) = − .
dy(x) y(x)
We get
q(x) = − log[y(x)] − σ0 .
Invert the above

1 dx 1
≡ =− ,
q(x) dy(x) log[y(x)] + σ0
multiply both sides by d y(x) and integrate

dy(x)
dx = − .
log[y(x)] + σ0
[Note it is helpful to know also the equality:

exp(log[y(x)]+σ0 )
− dy(x)
log[y(x)]+σ0
= − exp(−σ0 ) log[y(x)]+σ0
d(log[y(x)] + σ0 ).]
The result is

(c) : x + σ1 = − exp(−σ0 ) Ei log y(x) + σ0 . (7.22)
Solution (d)
(d) : D 2 y(x) + 2 y(x) . Dy(x) = 0 . (7.23)
Equation (d) can be shown to lead to
dq(x)
= −2 y(x) . (7.24)
dy(x)
Integration with respect to y gives

dq(x) dy(x)
. dy(x) = q(x) =
dy(x) dx

= − 2 y(x) . dy(x) = (σ0 )2 − y(x)2 (7.25)
where (σ0 )2 is a constant. Invert (7.25) and integrate with respect to y(x).

dx dy(x)
. dy(x) = dx = ,
dy(x) (σ0 )2 − y(x)2
1 y(x)
= x = tanh−1 − σ1 . (7.26)
σ0 σ0
Thus, the final result
(d) : y(x) = σ0 tanh [σ0 · (x + σ1 )] . (7.27)

Solve the following two differential equations (1) and (2) .
(1) : D 2 y + y Dy = 0 ;
(2) : D 2 y + exp(y) Dy = 0 . (7.28)
7.3 Order Reduction
Occasionally, the unlikely is true and one of the ν non-zero solutions to a νth order
homogeneous linear (ODE) is already known. When that happens, the given equation
can be reduced to an equation of the (ν − 1)th order.
Consider, for example, a νth order homogeneous linear (ODE)
ν

ai (x)D i u(x) = 0 (7.29)
i=0
where {ai (x)}, i = 0, 1, ..., ν are all known functions of x. The solution u(x) of
(7.29) is of course unknown. Given g(x) is a known solution of (7.29). Consider a
function φ(x) that is a product of g(x) and a new function f (x).
φ(x) = f (x) g(x) . (7.30)
Assume f (x) is so chosen that φ(x) is also a solution of (7.29). This require-
ment results in D f (x) satisfying homogeneous linear (ODE) of the (ν − 1)th order.
Because the size of the relevant algebra decreases rapidly with decrease in order, for
simplicity one works with a low value of ν. Accordingly, in the following, only a
second-order linear (ODE)is treated.

a0 (x) + a1 (x) D + a2 (x) D 2 u(x) = 0 . (7.31)
The two solutions to this equation are notated u 1 (x) and u 2 (x). Regarding these
solutions:
(a): Assume u 1 (x) is a known (or correctly guessed) non-trivial solution.

a0 (x) + a1 (x) D + a2 (x) D 2 u 1 (x) = 0 (7.32)
(b): The other linearly independent solution, u 2 (x) , to the same differential equation,
that is

a0 (x) + a1 (x) D + a2 (x) D 2 u 2 (x) = 0 , (7.33)
is currently unknown.
7.3 Order Reduction 203
7.3.1 Linear Independence of u1 and u2
It was asserted above that functions u 1 (x) and u 2 (x)—see (7.32) and
(7.33)—are linearly independent. It is helpful to show that this assertion is correct.
According to (3.47), two functions u 1 (x) and u 2 (x) = f (x) u 1 (x) are linearly
independent if and only if their Wronskian is non vanishing. In other words, functions
u 1 (x) and u 2 (x) are linearly independent only if the following is true

u 1 (x) u 2 (x) u 1 (x) f (x) u 1 (x)

Du 1 (x) Du 2 (x) = Du 1 (x) D[ f (x) u 1 (x)] = u 1 (x) · D f (x) = 0 .
2
(7.34)
With f (x), as in (7.45), the right-hand side of (7.34) is

a1 (x)
(1) : u 21 (x) · D f (x) = σ0 · exp − dx . (7.35)
a2 (x)
As such, the Wronskian is non-zero as long as a1 (x) is non-infinite and a2 (x) is non-
zero. Assuming both these requirements are satisfied, u 1 (x) and u 2 (x) are indeed
linearly independent.
7.4 Reduce Order from Second to First
Represent the second solution as a product of the first and a new function f (x). That is
u 2 (x) = f (x) u 1 (x) . (7.36)
Because u 1 (x) is already known, in order to determine u 2 (x) all one needs is calculate
f (x). And to that purpose, proceed as follows:
(A): Differentiate (7.36) twice.
Du 2 (x) = f (x) Du 1 (x) + u 1 (x) D f (x) ;

D 2 u 2 (x) = f (x) D 2 u 1 (x) + 2 Du 1 (x) D f (x) + u 1 (x) D 2 f (x) . (7.37)
(B): Substitute these differentials into (7.33).
0 = a0 f (x) u 1 (x) + a1 (x)[ f (x) Du 1 (x) + u 1 (x) D f (x)]

+a2 (x) [ f (x) D 2 u 1 (x) + 2 Du 1 (x) D f (x) + u 1 (x) D 2 f (x)] . (7.38)
(C): Rewrite (7.38). as

0 = f (x) a0 + a1 (x) D + a2 (x) D 2 u 1 (x) + a1 (x) u 1 (x) D f (x)
+ 2 a2 (x) Du 1 (x) D f (x) + a2 (x) u 1 (x) D 2 f (x) . (7.39)
As per (7.32), the first term on the right-hand side of (7.39) is equal to zero. By
introducing a convenient notation,
v(x) = D f (x) (7.40)
one can write the remaining parts of (7.39) as
[a1 (x) u 1 (x) + 2 a2 (x)Du 1 (x)] v(x) + a2 (x) u 1 (x) Dv(x) = 0. (7.41)
Dividing both sides by a2 (x)u 1 (x)v(x) and then multiplying by dx yields a neater
version of (7.41). That is
a1 (x) du 1 (x) dv(x)

dx + 2 + =0. (7.42)
a2 (x) u 1 (x) v(x)
Integration gives

a1 (x)
dx + 2 log[u 1 (x)] + log[v(x)] = log(σ0 )
a2 (x)
and exponentiating both sides leads to

a1 (x)
exp dx · [u 1 (x)]2 · v(x) = σ0 . (7.43)
a2 (x)
As such

a1 (x)
exp − a2 (x)
dx
v(x) = σ0 . (7.44)
[u 1 (x)]2
d f (x)
Because according to (7.40), v(x) = dx
, and therefore an integration with respect
to x would give

exp − a1 (x)
dx
a2 (x) d f (x)
v(x)dx = σ0 dx = dx = f (x) .
[u 1 (x)]2 dx
The conclusion: if one solution is known—that is, if u 1 (x) is already known—for

a second-order linear ordinary differential equation, one can always find its second
solution u 2 (x).

x exp − a1 (x)
dx
a2 (x)
(D) : u 2 (x) = u 1 (x) f (x) = σ0 u 1 (x) · dx . (7.45)
0 [u 1 (x)]2
7.4 Reduce Order from Second to First 205
7.4.1 Examples (I)–(VII)
The seven differential equations and their first solution, u 1 (x) , are given below. By
using the procedure ‘reduce order from second to first’ work out the second solu-
tion u 2 (x).
(I ) : (D 2 + 4)u(x) = 0 . : u 1 (x) = σ1 sin(2x)

(I I ) : (D 2 + b D − 2 b2 )u(x) = 0 . : u 1 (x) = σ1 exp(b x)
(I I I ) : [D 2 + D − 2]u(x) = 0 . : u 1 (x) = σ1 exp(x)
(I V ) : [D 2 − 2 D + 1]u(x) = 0 . : u 1 (x) = σ1 exp(x)
(V ) : (x 2 D 2 − x D − 3)u(x) = 0 . : u 1 (x) = σ1 x 3
x2
(V I ) : (D 2 + x D + 1)u(x) = 0 . : u 1 (x) = σ1 exp −
2
x2
(V I I ) : (D 2 + x D + x)u(x) = 0 . : u 1 (x) = σ1 exp x − (7.46)
2
7.4.2 Solution
(I ) : u 1 (x) = σ1 sin(2x) ; a2 (x) = 1 ; a1 (x) = 0 , a0 (x) = 4 ;

0
exp − dx
u 2 (x) = σ1 sin(2x) · σ0 1
dx
[σ1 sin(2x)]2

σ3
= sin(2x) · dx = σ2 cos(2x) = u 2 (x) : (1) .
[sin(2x)]2
(I I ) : u 1 (x) = σ1 exp(b x) ; a2 (x) = 1 ; a1 (x) = b , a0 (x) = −2b2 ;

b
exp − dx
u 2 (x) = σ1 exp(b x) · σ0 1
dx
[σ1 exp(b x)]2
= σ2 exp(−2 b x) = u 2 (x) . : (2) .
(I I I ) : u 1 (x) = σ1 exp(x) ; a2 (x) = 1 ; a1 (x) = 1 , a0 (x) = −2 ;

exp − 11 dx
u 2 (x) = σ1 exp(x) · σ0 dx
[exp(x)]2
= σ2 exp(−2 x) = u 2 (x) : (3) .
(I V ) : u 1 (x) = σ1 exp(x) ; a2 (x) = 1 ; a1 (x) = −2 , a0 (x) = 1 ;

exp (2)dx
u 2 (x) = σ1 exp(x) · σ0 dx
[exp(x)]2
= σ2 x exp(x) = u 2 (x) : (4) .
(V ) : u 1 (x) = σ1 x 3 ; a2 (x) = x 2 ; a1 (x) = −x , a0 (x) = −3 ;

−x
exp − x2
dx σ2
u 2 (x) = σ1 x · σ0
3
dx = : (5) .
[σ1 x ]
3 2 x
x2
(V I ) : u 1 (x) = σ1 exp − ; a2 (x) = 1 ; a1 (x) = x , a0 (x) = 1 ;
2

x2 exp − x1 dx
u 2 (x) = σ1 exp − · σ0 dx
2 [exp(−x 2 )]

x2 x2
= σ2 exp − · exp dx = u 2 (x) : (6) .
2 2
x2
(V I I ) : u 1 (x) = σ1 exp x − ; a2 (x) = 1 ; a1 (x) = x , a0 (x) = x ;
2

x2 exp (−x)dx
u 2 (x) = σ1 exp x − · σ0 dx
2 2
[exp x − x ]2 2
2 2
x x
= σ2 exp x − · exp − 2x dx = u 2 (x) : (7) .
2 2
(7.47)
Given one solution, u 1 (x), find the second solution, u 2 (x), of the following differ-
ential equations. [Hint: Read (7.32)–(7.47).]
(1) : (D 2 + 1)u(x) = 0 . : u 1 (x) = σ1 sin(x)

√
(2) : (D 2 + 2 D − 1)u(x) = 0 . : u 1 (x) = σ1 exp 2−1 x
1 x x
(3) : D2 + D + u(x) = 0 . : u 1 (x) = σ1 exp − sin
2 2 2
(4) : (D 2 − 2 D + 2)u(x) = 0 . : u 1 (x) = σ1 exp(x) sin(x)
√
(5) : (x 2 D 2 − x D − 1)u(x) = 0 . : u 1 (x) = σ1 x 1+ 2
x2
(6) : (D 2 − x D − 1)u(x) = 0 . : u 1 (x) = σ1 exp
2
x2
(7) : (D 2 + x D − x)u(x) = 0 . : u 1 (x) = σ1 exp −x − .
2
7.4.4 Particular Integral
Second-order homogeneous linear ordinary differential equations were treated in the

preceding section. Similar homogeneous linear (ODE)’s that are made inhomoge-
neous are treated here. The latter is done by the addition of a known function, B(x),
to the right-hand side in place of the zero that normally appears in homogeneous
linear (ODE)’s.
If Scs (x) is a complete solution of such equation—compare, (2.5)—one would
have

a0 + a1 (x) D + a2 (x) D 2 Scs (x) = B(x) . (7.48)
As noted in (3.59)—also compare (2.16)—the complete solution, Scs (x) is the sum
of the complementary solution, Scomp (x) ≡ u 1 (x) + u 2 (x) , and a particular integral,
I pi (x). That is
Scs (x) = Scomp (x) + I pi (x) = u 1 (x) + u 2 (x) + I pi (x) .
Thus, in addition to solving the differential equations

a0 + a1 (x) D + a2 (x) D 2 u 1 (x) = 0 ,

a0 + a1 (x) D + a2 (x) D 2 u 2 (x) = 0 , (7.49)
one would also need to solve

a0 + a1 (x) D + a2 (x) D 2 I pi (x) = B(x) . (7.50)
7.4.5 Calculation of I pi (x) Variation of Parameters
As per our previous experience determining I pi (x) for inhomogeneous linear (ODE)’s
with constant coefficients—see (3.60)–(3.91)—it is reasonable to expect the current
particular integral, I pi (x), will also bear relationship to the solution of the homoge-
neous linear (ODE) (7.49): namely u 1 (x) and u 2 (x). With this expectation in mind,
introduce a trial solution of the form
I pi (x) = f 1 (x) u 1 (x) + f 2 (x) u 2 (x) , (7.51)
where f 1 (x) and f 2 (x) are unknown functions that need to be determined. Clearly,
in order to determine f 1 (x) and f 2 (x) , two independent differential equations that
involve these functions are required.
Another important point to note is that the basic differential equation that specifies
I pi (x)—that is (7.50)—is second order. Therefore, one would need to differentiate
the trial solution (7.51) twice.
7.4.6 Procedure
Differentiate (7.51).
D I pi (x) = f 1 (x) Du 1 (x) + f 2 (x) Du 2 (x)

+ [u 1 (x) D f 1 (x) + u 2 (x) D f 2 (x)] . (7.52)
Differentiating once again would obviously introduce D 2 f 1 (x) and D 2 f 2 (x) . But
most certainly, unless it cannot be avoided, nobody wants to be stuck with having to
deal with second-order differential equations for the yet to be determined functions
f 1 (x) and f 2 (x) . Therefore, if at all possible, second differentials of f 1 (x) and f 2 (x)
must be avoided. Obviously, therefore, one must set that part of (7.52) equal to zero
which upon differentiation would introduce the unwanted second differentials. This
implies setting
{F I RST } : [u 1 (x) D f 1 (x) + u 2 (x) D f 2 (x)] = 0 . (7.53)
As a result, one now has
D I pi (x) = f 1 (x) Du 1 (x) + f 2 (x) Du 2 (x) ;

D 2 I pi (x) = D f 1 (x) Du 1 (x) + D f 2 (x) Du 2 (x)
+ f 1 (x) D 2 u 1 (x) + f 2 (x) D 2 u 2 (x) . (7.54)
Rewrite the original differential equation, namely (7.50), by making use of (7.51)
and (7.54). That is, write
B(x) = [a0 (x) + a1 (x)D + a2 (x)D 2 ]I pi (x) ;

= a0 (x)[ f 1 (x) u 1 (x) + f 2 (x) u 2 (x)]
+ a1 (x)[ f 1 (x) Du 1 (x) + f 2 (x) Du 2 (x)]
+ a2 (x) [D f 1 (x) Du 1 (x) + D f 2 (x) Du 2 (x) + f 1 (x) D 2 u 1 (x) + f 2 (x) D 2 u 2 (x)]
(7.55)
as
B(x) = f 1 (x)[a0 (x) + a1 (x)D + a2 (x)D 2 ]u 1 (x)

+ f 2 (x)[a0 (x) + a1 (x)D + a2 (x)D 2 ]u 2 (x)
+ a2 (x)[D f 1 (x) Du 1 (x) + D f 2 (x) Du 2 (x)] . (7.56)
Recall that each of the two functions u 1 (x) and u 2 (x) are solutions of the differ-
ential equation (7.49). As such, the top two lines on the right-hand side of (7.56) are
zero, and the third can be written as
{S EC O N D} : D f 1 (x) Du 1 (x) + D f 2 (x) Du 2 (x) = {B(x)/a2 (x)} . (7.57)
Clearly, (7.57) is the desired {S EC O N D} of the two differential equations—the first

being (7.53) named {F I RST }—that D f 1 (x) and D f 2 (x) must satisfy.
Equations {F I RST } and ({S EC O N D} are simultaneous linear equations in two
unknowns D f 1 (x) and D f 2 (x). Using elementary algebra, one finds
d f1 u 2 (x)
D f 1 (x) = = − {B(x)/a2 (x)} ;
dx u 1 (x)Du 2 (x) − u 2 (x)Du 1 (x)
d f2 u 1 (x)
D f 2 (x) = = {B(x)/a2 (x)} . (7.58)
dx u 1 (x)Du 2 (x) − u 2 (x)Du 1 (x)
Integration leads to the functions f 1 (x) and f 2 (x). Recall that one started off already
knowing the solution to the homogeneous part of the differential equation—namely
u 1 (x) and u 2 (x)—therefore the particular integral,
I pi (x) = f 1 (x) u 1 (x) + f 2 (x) u 2 (x),
can now be evaluated.
7.4.7 Examples: I pi (x) − (1) → I pi (x) − (4)
We determine the I pi (x) for examples numbered (1) - (4). Given alongside, these
equations are the relevant u 1 (x) and u 2 (x) : that is, both the first and the second
solution to the homogeneous part of each of these differential equations. Note that
these equations have constant coefficients. Therefore, they can be solved by using
‘undetermined coefficients’ that were described in detail in (3.60)–(3.91). Indeed, for
inhomogeneous linear (ODE)’s with constant coefficients, undetermined coefficients
are much easier to use than the ‘variation of parameters’ procedure currently under
discussion. Still, for demonstrational purposes, we use the more laborious variation
of parameters procedure described above.
I pi (x) − (1) : (D 2 + 4)u(x) = x exp(x) ;

u 1 (x) = σ1 sin(2x) ; u 2 (x) = σ2 cos(2x) . (1)
I pi (x) − (2) : (D + b D − 2 b )u(x) = exp(x) ;
2 2
u 1 (x) = σ1 exp(b x) ; u 2 (x) = σ2 exp(−2bx). (2)

I pi (x) − (3) : (D 2 + D − 2)u(x) = 1/x ;
u 1 (x) = σ1 exp(x) ; u 2 (x) = σ2 exp(−2x) . (3)
I pi (x) − (4) : (D 2 − 2 D + 1)u(x) = x ;
u 1 (x) = σ1 exp(x) ; u 2 (x) = σ2 x exp(x) . (4) (7.59)
7.4.8 Solution
Employe (7.58), calculate D f 1 (x) and D f 2 (x), integrate the result, and determine
f 1 (x) and f 2 (x).
x
I pi (x) − (1) : D f 1 (x) = exp(x) cos(x) ,
2σ1
x
D f 2 (x) = − exp(x) sin(x) ,
2σ2
exp(x)
f 1 (x) = [(10x − 4) sin(2x) + (5x + 3) cos(2x)] + σ3 ,
50σ1
exp(x)
f 2 (x) = − [(5x + 3) sin(2x) + (4 − 10x) cos(2x)] + σ4 .
50σ2
Given u 1 (x) = σ1 sin(2x) and u 2 (x) = σ2 cos(2x)—see equation I pi (x) − (1)

in Examples I pi (x) —the relevant I pi (x) is readily found.
I pi (x) − (1) : I pi (x) = f 1 (x) u 1 (x) + f 2 (x) u 2 (x)

x exp(x) 2 exp(x)
= − + σ3 u 1 (x) + σ4 u 2 (x) .
5 25
1
I pi (x) − (2) : D f 1 (x) = exp[(1 − b)x] ,
3 b σ1
1
D f 2 (x) = − exp[(1 + 2 b)x] ,
3 b σ2
1 1
f 1 (x) = exp[(1 − b)x] + σ3
3 b σ1 1−b
1 1
f 2 (x) = − exp[(1 + 2 b)x] + σ4 .
3 b σ2 1+2b
Given u 1 (x) = σ1 exp(b x) and u 2 (x) = σ2 exp(−2 b x)—see equation

I pi (x) − (2) in examples I pi (x) —the relevant I pi (x) is

exp(x)
= 3b
1
1−b
− 1+2b1
+ σ3 u 1 (x) + σ4 u 2 (x) .
1 exp(x)
I pi (x) − (3) : D f 1 (x) = ,
3 σ1 x
1 exp(2 x)
D f 2 (x) = − ,
3 σ2 x
1
f 1 (x) = Ei(−x) + σ3
3 σ1
1
f 2 (x) = − Ei(2 x) + σ4 .
3 σ2
Given u 1 (x) = σ1 exp(x) and u 2 (x) = σ2 exp(−2 x) ,—see equation

I pi (x) − (3) in examples I pi (x) — the relevant I pi (x) is

exp(x) 1
= Ei(−x) − exp(−2 x) Ei(2 x) + σ3 u 1 (x) + σ4 u 2 (x) .
3 3
Given u 1 (x) = σ1 exp(x) and u 2 (x) = σ2 x exp(x)—see equation I pi (x) − (4)
in examples I pi (x) —the relevant I pi (x) is found by following a similar procedure
to that used above.
1
I pi (x) − (4) : D f 1 (x) = − x 2 exp(−x) ,
σ1
1
D f 2 (x) = − x exp(−x) ,
σ2
1
f 1 (x) = (x 2 + 2x + 2) exp(−x) + σ3
σ1
1
f 2 (x) = − (x + 1) exp(−x) + σ4 .
σ2
Therefore, one has

= (x 2 + 2x + 2) − x(x + 1) + σ3 u 1 (x) + σ4 u 2 (x) .
= x + 2 + σ3 u 1 (x) + σ4 u 2 (x) (7.60)
7.4.9 Examples: I pi (x) − (5) → I pi (x) − (10)
Somewhat more complicated equations than those numbered (1)–(4) are treated
below. Note: given alongside, these differential equations are u 1 (x) and u 2 (x) , which
are the two solutions to the homogeneous part of each of these differential equations.
1 σ
2
(5) (x 2 D 2 − x D − 3)u(x) = ; u 1 (x) = σ1 x 3 ; u 2 (x) =
x2 x
(6) (x 2 D 2 + x D + 1)u(x) = log(x) ;
u 1 (x) = σ1 sin[log(x)] ; u 2 (x) = σ2 cos[log(x)]
(7) [x 2 D 2 + x D + 4]u(x) = x 2 log(x) ;

u 1 (x) = σ1 sin 2 log(x) ; u 2 (x) = σ2 cos 2 log(x)
1
(8) [x(x + 1) D 2 − (x + 1) D + 1]u(x) = ; u 1 (x) = σ1 (x + 1) ;
x
u 2 (x) = σ2 [x log(x + 1) + log(x + 1) + 1]
(9) [(x 2 − 1) D 2 − (x 2 + 2x − 1) D + 2x]u(x) = x ; u 1 (x) = σ1 (x + 1)2 ;
u 2 (x) = σ2 exp(x) :
(10) [(x + 2) D + (x + 2) D + 1]u(x) = (x + 2) ;
2 2

u 1 (x) = σ1 sin log(x + 2) ; u 2 (x) = σ2 cos log(x + 2) . (7.61)
These differential equations, namely (7.61)-(5)–(7.61)-(10), do not have constant

coefficients. Therefore, for determining their I pi (x) , it is appropriate to use the
‘variation of parameters’ procedure. Recall that the variation of parameters procedure
was discussed in detail when equations (1)–(4) were solved above.
7.4.10 Solution
As before, use (7.58); calculate D f 1 (x) and D f 2 (x); integrate the result; and deter-
mine f 1 (x) and f 2 (x).
−1
I pi (x) − (5) : f 1 (x) = − 20σ1 x 5 + σ3 ; f 2 (x) = (4σ2 x)−1 + σ4 .
σ2
Using the information: u 1 (x) = σ1 x 3 , u 2 (x) = x
and
I pi (x) − (5) = f 1 (x) u 1 (x) + f 2 (x) u 2 (x) ,
one gets
1 1 x −2
I pi (x) − (5) : I pi (x) = − + = + σ3 u 1 (x) + σ4 u 2 (x) .
20 x 2 4 x2 5
Next, consider equation I pi (x) − (6) .
1
I pi (x) − (6) : f 1 (x) = {log(x) sin[log(x)] + cos[log(x)]} + σ3 ;
σ1
1
f 2 (x) = − {sin[log(x] − log(x) cos[log(x)]} + σ4 .
σ2
And the fact that u 1 (x) = σ1 sin[log(x)] , u 2 (x) = σ2 cos[log(x)] and
I pi (x) − (6) = f 1 (x) u 1 (x) + f 2 (x) u 2 (x)
leads to the result
I pi (x) − (6) : I pi (x) = log(x) + σ3 u 1 (x) + σ4 u 2 (x) .
In similar fashion equation I pi (x) − (7) leads to the following:
x2
I pi (x) − (7) : f 1 (x) = {[2 log(x) − 1] sin[2 log(x)] + 2 log(x) cos[2 log(x)]} + σ3 ;
16σ1
x2
f 2 (x) = {[2 log(x) − 1] cos[2 log(x)] − 2 log(x) sin[2 log(x)]} + σ4 .
16σ2

Therefore, using u 1 (x) = σ1 sin 2 log(x) , u 2 (x) = σ2 cos 2 log(x) and the
equation
I pi (x) − (7) = f 1 (x) u 1 (x) + f 2 (x) u 2 (x) ,
one is led to
x2 x2
I pi (x) − (7) : I pi (x) = log(x) − + σ3 u 1 (x) + σ4 u 2 (x) .
8 16
Proceeding as in equation I pi (x) − (6) one has

1 x2 + 1 log(x) 1 − x
(8) : f 1 (x) = log(x + 1) − + + σ3 ;
σ1 2x 2 2 2x 2
1 1
f 2 (x) = − + σ4 .
σ2 2x 2

Therefore, using u 1 (x) = σ1 sin 2 log(x) , u 2 (x) = σ2 cos 2 log(x) and the
equation
I pi (x) = f 1 (x) u 1 (x) + f 2 (x) u 2 (x)

gives
1
(7.61) − (8) : I pi (x) = (x + 1) log(x + 1) − (x + 1) log(x) − 1
2
+ σ3 u 1 (x) + σ4 u 2 (x) .
Proceeding as usual, (7.61)-(9) leads to
1 1
(9) : f 1 (x) = + σ3 ;
2σ1 (x 2 − 1)
1 exp(−x)
f 2 (x) = − + σ4 .
σ2 x −1
Therefore, using u 1 (x) = σ1 (x + 1)2 , u 2 (x) = σ2 exp(x) and the relationship
I pi (x) = f 1 (x) u 1 (x) + f 2 (x) u 2 (x)
one gets
1
(7.61) − (9) : I pi (x) = + σ3 u 1 (x) + σ4 u 2 (x) .
2
Finally, we treat (7.61)-(10).
1 x +2
(10) : f 1 (x) = {sin[log(x + 2)] + cos[log(x + 2)]} + σ3 ;
σ1 2
1 x +2
f 2 (x) = − {sin[log(x + 2)] − cos[log(x + 2)]} + σ4 .
σ2 2
Therefore, using u 1 (x) = σ1 sin[log(x + 2)] , u 2 (x) = σ2 cos[log(x + 2)] and the
relationship
I pi (x) = f 1 (x) u 1 (x) + f 2 (x) u 2 (x)
one readily finds

x
(7.61) − (5) : I pi (x) = + 1 + σ3 u 1 (x) + σ4 u 2 (x) .
2
Note: Because σ3 u 1 (x) + σ4 u 2 (x) are already present in the complementary
solution, they can be ignored from the relevant result for I pi (x) for differential equa-
tions (7.59)-(1) → (7.59)-(10).
Determine the I pi (x) for the following inhomogeneous linear ordinary differential
equation with constant coefficients numbered (1)–(10). Given alongside are u 1 (x)
and u 2 (x) : that is, both the first and the second solution to the homogeneous part of
each of these differential equations. [Hint: Read (7.48)–(7.58).]
(1) (D 2 + 1)u(x) = exp(x) ; u 1 (x) = σ1 sin(x) ; u 2 (x) = σ2 cos(x) .

√
(2) (D 2 + 2 D − 2)u(x) = x 2 ; u 1 (x) = σ1 exp 3−1 x ;
√
u 2 (x) = σ2 exp − 3 − 1 x : (2)
√
3 x 5
(3) D + D+
2
u(x) = x ; u 1 (x) = σ1 exp − sin x ;
2 2 2
√
x 5
u 2 (x) = σ2 exp − cos x : (3)
2 2
√
√ 3
(4) (x D + 1)u(x) = log(x) ; u 1 (x) = σ1 x sin
2 2
log(x) ;
2
√
√ 3
u 2 (x) = σ2 x cos log(x) : (4)
2
1 √ √
(5) (x 2 D 2 − x D − 1)u(x) = ; u 1 (x) = σ1 x 1+ 2 ; u 2 (x) = σ2 x 1− 2 .
x
(6) (x 2 D 2 + x D + 1)u(x) = x log(x) ;
u 1 (x) = σ1 sin[log(x)] ; u 2 (x) = σ2 cos[log(x)] : (6)
(7) (x D + x D + 3)u(x) = (1/x) log(x) ;
2 2
√ √
u 1 (x) = σ1 sin 3 log(x) ; u 2 (x) = σ2 cos 3 log(x) ; (7)
2
(8) [x(x + 1) D 2 − (x + 1) D + 1]u(x) = ; u 1 (x) = σ1 (x + 1) ;
x
u 2 (x) = σ2 [x log(x + 1) + log(x + 1) + 1] : (8)
(9) [(x 2 + 2x) D 2 − (x 2 + 4x + 2) D + 2(x + 1)]u(x) = x + 1 ;
u 1 (x) = σ1 (x + 2)2 ; u 2 (x) = σ2 exp(x) : (9)
(10) [(x + 1) D + (x + 1) D + 1]u(x) = (x + 1) ;
2 2

u 1 (x) = σ1 sin log(x + 1) ; u 2 (x) = σ2 cos log(x + 1) .(10) (7.62)
7.5 Other Special Situations
Consider the following second-order linear (ODE) with variable coefficients.

M0 (x) + M1 (x) D + D 2 u(x) = B(x) . (7.63)
The complete solution is the sum of complementary solution and the particular
integral. The complementary solution of a second-order differential equation with
variable coefficients can be very hard to find. Generally, it is constituted of two non-
trivial solutions, and assuming neither of these can conveniently be found, one looks
for special situations. And there are some such situations, but they require appropriate
transformations that use either M0 (x) or M1 (x), functions that occur in (7.63).
7.6 Transformation Using M1 (x) (7.63)
In an attempt to solve (7.63), try a transformation that utilizes M1 (x). Represent the
solution of (7.63), namely u(x), as
u(x) = y(x) H (x) , (7.64)
where

H (x) = exp α M1 (x) dx . (7.65)
Inserting (7.64) into (7.63) and doing a little bit of algebra, one can rewrite the latter
as

α (D M1 ) + (M1 )2 (α2 + α) + M0 + M1 (2α + 1) D + D 2 y(x)
= H −1 (x) B(x) . (7.66)
If one sets α = − 21 , that is, if one uses the relationship
(2α + 1) = 0 , (7.67)
the term that multiplies D in (7.66) vanishes. And as a result, except for D 2 , the only
remaining term on the left-hand side of (7.66) is

α (D M1 ) + (M1 )2 (α2 + α) + M0

1 1
= − D M1 − M12 + M0 ≡ G . (7.68)
2 4
And a happy circumstance is that when G is either equal to a constant, or is propor-

tional to x −2 . That is, when
7.6 Transformation Using M1 (x) (7.63) 217
Either G=C , : (α)

−2
Or G=Cx , : (β) (7.69)
where C is a constant.
specifies y(x) =
Examine precisely what all this implies. Recall also that (7.64)
H −1 u(x) and with the use of (7.67), (7.65) gives H −1 = exp 21 M1 (x) dx . There-
fore, if in addition to (7.67) one uses also (7.69)-(α) , the differential equation for
y(x) , namely (7.66), becomes

C + D 2 y(x) = H −1 (x) B(x) , (7.70)
On the other hand, when one uses (7.69)-(β) the differential equation (7.66) leads to
−2
C x + D 2 y(x) = H −1 (x) B(x) . (7.71)
Note, both these equations for y(x)—namely (7.70) and (7.71)—would be much
easier to solve than the original differential equation (7.63).
A set of second-order differential equations are worked out below.

Reminder: These equations are written in the form of (7.63). So their solution is
represented as u(x). By using, (7.63)–(7.71) one calculates first the function y(x).
And becauseu(x) = y(x) H (x), after
this calculation
one multiplies the result by
H (x) = exp α M1 (x) dx = exp − 21 M1 (x) dx where α = − 21 .
2
(1) (4x − 2) − 4x D + D 2 u = exp(x 2 ) x .
2 x 2 + 2x
(2) x − 1 − (2x) D + D 2 u = x exp .
2
x2 x x 2
(3) − x D + D2 u = exp .
4 8 4
2 2
(4) 2
− D + D 2 u = x exp(x) .
x x

(5) 4(x − 1)2 − 4(x − 1)D + D 2 u = exp(x 2 ) .
2 2
(6) − D + D2 u = x .
(x + 2) 2 x +2
1 1
(7) − 2+ D + D 2 u = x −2 .
4x x

(8) 4 exp(2 x) − {1 + 4 exp(x)} D + D 2 u = exp(2 x) .
1 2
(9) − + D + D 2 u = x −1 .
x2 x
4 4
(10) − D + D2 u = x 3 . (7.72)
x2 x
7.6.2 Solution

(1) M0 = 4x 2 − 2 ; M1 = −4 x ; G = C = 0 ; H = exp(x 2 ) ;
B = exp(x 2 ) x .
Therefore,

C + D 2 y = 0 + D 2 y = H −1 B
= D2 y = x . (7.73)
This is a simple differential equation that can directly be integrated.
x3
y(x) = σ1 x + σ2 + . (7.74)
6
Therefore, the solution of the differential equation (7.72)-(1) is
x3
(1) u(x) = y H = y exp(x 2 ) = σ1 x + σ2 + exp(x 2 ) . (7.75)
6
x2
(2) M0 = x 2 − 1 ; M1 = −2 x ; G = C = 0 ; H = exp ;
2
x + 2x
2
B = x exp ; D 2 y = H −1 B = x exp(x) . (7.76)
2
Again, one can proceed by direct integration.
y = σ1 x + σ2 + (x − 2) exp(x) .
And the solution of (7.72)-(2) is:

x2 x2
(2) u = y H = y exp = {σ1 x + σ2 + (x − 2) exp(x)} exp .
2 2
(7.77)
x2 1 x2
(3) M0 = ; M1 = −x ; G = C = ; H = exp ;
4 2 4
x2 x 1 x
B = exp ; + D 2 y = H −1 G = . (7.78)
4 8 2 8
Differential equation (7.78) is both simple, has constant coefficients, and can readily
be solved by using the technique described earlier in (3.5)–(3.73), etc.
x x x
y = σ1 sin √ + σ2 cos √ + . (7.79)
2 2 4
The solution is:

x x x x2
(3) u = y H = σ1 sin √ + σ2 cos √ + exp . (7.80)
2 2 4 4
2 2
(4) M0 = 2
; M1 = − ; G = C = 0 ; H = x ;
x x
B = exp (x) x ; D 2 y = H −1 B = exp(x) . (7.81)
Again, one can directly integrate this equation.
y = σ1 x + σ2 + exp(x). (7.82)
And the solution is:
(4) u = y H = {σ1 x + σ2 + exp(x)} x . (7.83)
(5) M0 = 4(x − 1)2 ; M1 = −4(x − 1) ; G = C = 2 ; H = exp(x 2 − 2x) ;

B = exp x 2 ; [2 + D 2 ]y = H −1 B = exp(2x) .
As before, using the procedure described in (3.5)–(3.73) etc., one gets

√ √ 1
y = σ1 sin x 2 + σ2 cos x 2 + exp(2x). (7.84)
6
The solution is
(5) u = y H
√ √ 1
= σ1 sin x 2 + σ2 cos x 2 + exp(2x) exp(x 2 − 2x)
6
√ √ 1
= σ1 sin x 2 + σ2 cos x 2 exp(x 2 − 2x) + exp(x 2 ) .
6
(7.85)
2 2
(6) M0 = ; M1 = − ; G = C = 0 ; H = (x + 2) ;
(x + 2) 2 (x + 2)
x
B = x ; D 2 y = H −1 B = . (7.86)
x +2
Direct integration yields
x2
y = σ1 x + σ2 + − 2(x + 2) log(x + 2). (7.87)
2
Therefore,
x3
(6) u = y H = σ3 x + σ4 x 2 + − 2(x + 2)2 log(x + 2) . (7.88)
2
1 1
; G = C = 0 ; H = (x)− 2 ;
1
(7) M0 = − ; M1 =
4 x2 x
B = x −2 ; D 2 y = H −1 B = x − 2 .
3
(7.89)
Direct integration leads to

√
y = σ1 x + σ2 − 4 x . (7.90)
Hence, the solution:

√
(7) u = y H = [σ1 x + σ2 − 4 x](x)− 2 = σ1 (x) 2 + σ2 (x)− 2 − 4 .
1 1 1
(7.91)
−1
(8) M0 = 4 exp(2 x) ; M1 = {1 + 4 exp(x)} ; G = C = ;
x 4
H = exp + 2 exp(x) ; B = exp(2x) ;
2
−1 x
+ D 2 y = H −1 B = exp 3 − 2 exp(x) . (7.92)
4 2
Direct integration gives
1 x
y = σ1 exp(x/2) + σ2 exp(−x/2) + exp [− − 2 exp(x)] . (7.93)
4 2
And the solution is
1
(8) u = y H = σ1 exp x + 2 exp(x) + σ2 exp 2 exp(x) + .
4
(7.94)
1 2 −1
(9) M0 = − ; M1 = ; G=C = ;
x2 x x2
1 1 −1
H = ; B = ; + D 2 y = H −1 B = 1 . (7.95)
x x x2
As before, using the procedure described in (3.5)–(3.73) etc., one gets

√ √
5+1 − 5+1
y = σ1 x 2 + σ2 x 2 + x2 . (7.96)
Therefore, solution to the second-order differential equation (7.72)-(9) with variable

coefficients is the following.
√ √
5−1 − 5−1
(9) u = y H = σ1 x 2 + σ2 x 2 +x . (7.97)
4 4 −2
(10) M0 = ; M1 = − ; G=C = ;
x2 x x2
−2
H = x2 ; B = x3 ; + D 2 y = H −1 B = x . (7.98)
x2
Finally, one arrives at
σ2 x3
y = σ1 x 2 + + . (7.99)
x 4
Accordingly, the solution of differential equation (7.72)-(10) is
σ2 x3 2
(10) u = y H = σ1 x 2 + + x . (7.100)
x 4
Use the procedure outlined in (7.64)–(7.71) and determine the differential equation
obeyed by y(x). Solve it and calculate u(x).
2
(1) (4x + 3) − 4x D + D 2 u = exp(x 2 ) .
x2
(2) x 2 − 2x D + D 2 u = exp .
2
9 2 3 2
(3) x − 3x D + D 2 u = exp x .
4 4
1 2
(4) 2
− D + D 2 u = x −2 .
x x
7.7 Transformation Using M0 (x)
Recall (7.63) and (7.64). They are

M0 (x) + M1 (x) D + D 2 u(x) = B(x) ,
and , u(x) = y(x) H (x) . (7.101)
Unlike (7.65), here H (x) has been set equal to 1 so that
u(x) ≡ y(x) ,

and , M0 (x) + M1 (x) D + D 2 y(x) = B(x) . (7.102)
Let us introduce a new independent variable, z, defined by the relationship
dz
= (±)M0 . (7.103)
dx
√
In order that (±)M0 is real, the sign (±) is so chosen that (±)M0 is positive.
Consider the following relationships.
dy dy dz
Dy = = ; (7.104)
dx dz dx
7.7 Transformation Using M0 (x) 223

d d dy dz d dy dz dz
D2 y = [Dy] = =
dx dx dz dx dz dz dx dx
2
d y dz dy d dz dz
= + ·
dz 2 dx dz dz dx dx
2
d2 y dz dy d dz dz
= + ·
dz 2 dx dz dz dx dx
2
d2 y dz dy d dz
= + ·
dz 2 dx dz dx dx
2
d2 y dz dy d2z
= + · . (7.105)
dz 2 dx dz dx 2
Inserting this information into (7.102) leads to
dy d2 y B(x)
(±)y + K (±) + 2 = , (7.106)
dz dz (±)M0
where
d √(±)M √
0
+ M1 (±)M0
K (±) = dx
. (7.107)
(±)M0
Should K (±) turn out to be equal to a constant then (7.106) would be a second-order
linear ordinary differential equation with constant coefficients and as such would be
easy to solve. And, indeed, if this constant should happen to be zero, the solution
would be even easier to obtain.
7.8 Examples Group IX
Work out the following second-order differential equations.

(1) −cos2 (x) + tan(x) D + D 2 u = − sin(x) cos2 (x) .

(2) 12 exp(4x) − 2 (1 + 4 exp(2x)) D + D 2 u = 4 exp (2 exp(2x) + 4x) .
2 2
(3) 4
+ D + D 2 u = 2 x −4 + x −6 .
x x
1 1
(4) − 2
+ D + D2 u = x . (7.108)
x x
7.8.1 Solution
Equation (7.108)-(1)
In (7.108)-(1), currently, we have M0 = −cos2 (x), but instead of M0 we choose
(±)M0 which is required to be +cos2 (x). Therefore, in (1) one must set (±) = − 1.
One follows the same procedure in equations (2) → (4).
H er e , (±) ≡ (−) .
dz
Also : = (±)M0 = cos(x) ; z = sin(x) ; M1 = tan(x) .
dx
√
d (±)M0
= − sin(x) ; M1 (±)M0 = sin(x) ; K (±) = 0 ;
dx
B(x) = − sin(x) cos2 (x) . (7.109)
Accordingly, via the use of (7.103)–(7.107), (7.108)-(1) is reduced to (±)y +

d2 y
dz 2
= (±)M
B(x)
0
. Or equivalently
d2 y − sin(x) cos2 (x)

−y+ = = − sin(x)
dz 2 cos2 (x)
= −z . (7.110)
This equation has constant coefficients and is readily solved by using the techniques
described in a previous chapter. Its solution is:
(1) : y = σ1 exp(z) + σ2 exp(−z) + z . (7.111)
One can change back to the original variable x by substituting sin(x) for z. Therefore,
the solution of (7.108)-(1) is:
(1) : y = σ1 exp[sin(x)] + σ2 exp[− sin(x)] + sin(x) . (7.112)
In (7.108)-(2), M0 is positive. So one can safely ignore the use of the symbol (±).
dz √ √
(2) M0 = 12 exp(4x) ; =
M0 = 12 exp(2x) ; z = 3 exp(2x) ;
dx
√
d M0 √
M1 = − 2[1 + 4 exp(2x)] ; = 48 exp(2x) ;
dx
√ 4
M1 M0 = − 48 [exp(2x) + 4 exp(4x)] ; K = − √ ;
3
B(x) 1 1 2z
= exp 2 exp(2x) = exp √ . (7.113)
M0 3 3 3
7.8 Examples Group IX 225
Accordingly, via the use of (7.103)–(7.106), (7.108)-(2) is reduced to
4 dy d2 y B(x) 1 2z
y− √ + 2 = = exp √ . (7.114)
3 dz dz M0 3 3
This equation has constant coefficients and is readily solved by using the techniques
described in a previous chapter. Its solution is:
z √ 2z
(2) : y = σ1 exp √ + σ2 exp z 3 − √ . (7.115)
3 3
√
One can change back to the original variable x by substituting 3 exp(2x) for z.
Therefore, the solution of (7.108)-(2) is
(2) : y = σ1 exp[exp(2x)] + σ2 exp[3 exp(2x)] − exp[2 exp(2x)] .(7.116)
(3): Next consider (7.108)-(3). Here too, M0 is positive. So the symbol (±) is not
needed.
√ √
2 dz 2 2
(3) : M0 = 4 ; = M0 = 2 ; z = − ;
x dx x x
√ √
2 d M0 8
M1 = ; = − 3 ;
x dx x
√ √
8 2
M1 M0 = 3 ; K = 0 ; x = − .
x z
B 2 (x −4 + x −6 ) −2 z2
= = 1 + x = 1 + . (7.117)
M0 2(x −4 ) 2
Accordingly, via the use of (7.103)–(7.106), (7.108)-3 is reduced to
d2 y B z2
y+ 2
= = 1+ . (7.118)
dz M0 2
This equation is readily solved.
z2
(3) : y = − σ1 sin(z) + σ2 cos(z) + . (7.119)
2
√
Again, one can change back to the original variable by substituting − z2 for z.
And the solution of (7.108)-(3) is
√ √
2 2 1
(3) : y(x) = σ1 sin + σ2 cos( + 2 . (7.120)
x x x
−1
(4): In (7.108)-(4), M0 is equal to x2
. Therefore, as suggested earlier, we choose
(±)≡ − .
1 dz 1
(4) : (±)M0 = 2
; = (±)M0 = ; T her e f or e, z = log x ;
x dx x
√
1 d (± )M0 1
Also, M1 = ; = − 2 ;
x dx x
1
M1 ± M0 = 2 ; K (±) = 0 ; B = x . (7.121)
x
Accordingly, via the use of (7.103)–(7.107), (7.108)-(4) is reduced to
d2 y B
(±)y + = ,
dz 2 (±)M0
d2 y x
−y + 2 = 1 = x 3 = exp(3 z) . (7.122)
dz x2
This equation has constant coefficient and is solved by previously described tech-
niques. Its solution is
exp(3 z)
y = σ1 exp(z) + σ2 exp(−z) + . (7.123)
8
Now change back to the original variable x by substituting log(x) for z in (7.123).
The result is the desired solution to (7.108)-(4).
σ2 x3
(4) : y(x) = σ1 x + + exp(−z) + . (7.124)
x 8
Chapter 8
Oscillatory Motion
Oscillatory motion is central to the description of acoustics and the effects of inter-
particle interaction in many physical systems. In its most accessible form, oscillatory
motion is simple harmonic. Such motion—which in this chapter is described first—
has a long and distinguished history of use in modeling physical phenomena .
Described next is anharmonic motion which somewhat more realistically repre-
sents the observed behavior of oscillatory physical systems. To this end, a detailed
analysis of transient state motion is presented for a point mass for two different
oscillatory systems. These are as follows:
(1) The point mass, m, is tied to the right end of a long, massless coil spring placed
horizontally in the x-direction on top of a long, level table. The left end of the
coil is fixed to the left end of the long table. The motion of the mass is slowed by
frictional force, that is, proportional to its momentum m v(t). In its completely
relaxed state, the spring is in equilibrium and the mass is in its equilibrium
position (EP).
(2) Because the differential equations needed for analyzing damped oscillating pen-
dula are prototypical of those used in the studies of electromagnetism, acoustics,
mechanics, chemical and biological sciences, and engineering, we analyze next
a pendulum consisting of a (point-sized) bob of mass m, that is, tied to the end of
a massless stiff rod of length l. The rod hangs down, in the negative z-direction,
from a hook that has been nailed to the ceiling. The pendulum is set to oscillate
in two-dimensional motion in the x − z plane. Air resistance is approximated as
a frictional force proportional to the speed with which the bob is moving. The
ensuing friction slows the oscillatory motion.

https://doi.org/10.1007/978-3-319-76406-1_8
228 8 Oscillatory Motion
8.1 Periodic Functions
A function F(x) is periodic if for all values of x there exists a positive constant,
CONST, such that
F(x + CONST) = F(x) . (8.1)
The period of the function is the smallest positive value of CONST, say, equal to 2P
so that
F(x + 2P) = F(x) . (8.2)
Beauty and simplicity of oscillatory periodic motion has long attracted scientists
interested in modeling observed physical phenomena.
The simplest possible periodic motion is called simple harmonic motion(s.h).
(s.h) is force-free, undamped periodic motion with constant period of oscillation and
unchanging size of the maximum and the minimum displacements.
Oscillatory Motion of Mass Tied to Spring on a Table
Consider, for instance, a point mass, m, tied to the right end of a long massless coil
spring placed horizontally in the x-direction on top of a long, level, friction-free
smooth table. The left end of the coil is fixed to the left end of the long table. In its
completely relaxed state, the spring is in equilibrium and the mass, tied to the right
end of the coil, is in its equilibrium position. The equilibrium position is henceforth
to be referred to as the (EP).
Pull the mass away from the (EP) by distance + x. The extension of the spring
beyond the (EP) provides a restoring force. In general, the strength of the restoring
force, F(x), is a complicated function of the extension x. If F(x) is analytic within
−X 0 ≤ x ≤ X 0 , the Maclaurin − T aylor 27.−28. series expansion obtains.
∞ i
dF
F(x) = F(0) + xi . (8.3)
i=1
dx i
Because there is no restoring force at zero extension, F(0) must be vanishing. Also,
for small extension |x| < |X 0 |— where |X 0 | is the maximum extension where the
Hooke’s law accurately holds—the Hooke’ law 29. applies. That is
F(x) ≈ −(m K ) x (8.4)

where (m K ) = − dF dx
is the so-called Hooke’s constant. If the spring is extended
toward the positive direction—meaning, if x is positive—the restoring force must be
toward the negative direction. Thus the Hooke’s constant (m K ) is positive.
8.1 Periodic Functions 229
In practice, the restoring force may not be linear. As such, there would be anhar-
monicity related to nonlinear contribution to F(x). Also, many oscillatory processes
encounter damping due to a variety of frictional effects. Yet, the restoring force F(x),
for moderate to small displacement, is often well approximated by Hooke’s law.
8.1.1 Harmonic Oscillation
The Newtonian equation of motion of the mass m that itself is constant—meaning,

dm
dt
= 0—and experiences a force F(x) is
d2 x
F(x) = m = m D2 x . (8.5)
dt 2
Using (8.4) and (8.5), the resultant second-order differential equation (relevant only
for small |x|) is
m [D 2 + 0 2 ]x = 0 , (8.6)
where
d dn
D = ; Dn = ; 0 2 = K . (8.7)
dx dx n
(Note: 0 is real because K is positive.)
The second-order differential equation (8.6) has constant coefficients. Therefore,
it is readily solved by using the standard rules described in (3.5) to (3.42). The
solution is sinusoidal
x(t) = σ0 cos (0 t + φ0 ) (8.8)
and has three parameters: σ0 , 0 , and φ0 . To determine them, proceed as follows.

The maximum value of cos(...) is = 1 making σ0 the maximum displacement of
the point mass beyond the (EP).
The period of the sinusoidal oscillatory harmonic motion is defined to be the
minimum time, τ , and it takes to exactly complete one whole cycle. That is
x(t + τ ) = x(t)
so that
cos [0 (t + τ ) + φ0 ] = cos (0 t + φ0 ) . (8.9)

This requires 0 τ to equal 2π.

2π 2π
τ = = √ . (8.10)
0 K
The frequency, ν, of the sinusoidal oscillation is
1
ν= . (8.11)
τ
The angular frequency of the oscillations is defined to be 2πν .
2π √
2πν = = 0 = K . (8.12)
τ
The one remaining parameter, namely the phase angle φ0 , can be determined from
one additional piece of information, e.g., either the position or the velocity at some
specific time. For instance, assume the mass is at its maximum displacement—that
is, σ0 —at time t = 0. Then, according to (8.8)
cos (φ0 ) = 1 (8.13)
meaning φ0 = 0. One could have got the same result also from requiring at maximum
displacement, σ0 , the velocity is vanishing. That is, at t = 0

dx
= − σ0 ω0 sin [ω0 t + φ0 ] = − σ0 ω0 sin [φ0 ] = 0 . (8.14)
dt t=0
Thus, φ0 is indeed zero, and the solution of the differential equation (8.6) is
x(t) = σ0 cos (0 t) . (8.15)
8.1.2 Energy
The (s.h) oscillatory motion described above is infinitely long-lived. How does the
system energy fair? The total energy is the sum of the kinetic energy
2
m dx m 1
E kinetic ≡ = [−x0 0 sin (0 t)]2 = m K (x0 )2 [sin (0 t)]2 ,
2 dt 2 2
and the potential energy, which is equal to the work done in extending the spring a
distance x
x
1 1
E potential ≡ − F(x) dx = m K x 2 = m K (x0 )2 [cos (0 t)]2 .
0 2 2
That is
1
E total = E kinetic + E potential = m K (x0 )2 [sin (0 t)]2 + [cos (0 t)]2
2
1
= m K (x0 )2 . (8.16)
2
Clearly, the total energy is constant, independent of time. Notice how at maximum
displacement in either direction when [cos (0 t)]2 = 1 and [sin (0 t)]2 = 0 energy
is all potential. This happens when 0 t = n π, n = 0, 1, 2, ... At the midpoint, that
is, at the (EP), when [cos (0 t)]2 = 0 and [sin (0 t)]2 = 1 the energy is all kinetic.
This happens whenever 0 t = (2 n + 1) (π/2). In between these two extremes, the
energy is partially kinetic and partially potential.
8.1.3 Energy Conservation and Equation of Motion
Conservation of total energy provides a convenient tool for determining the differ-
ential equation obeyed by the harmonically oscillating mass. Both the velocity, v,
and the position, x, in the expression for total energy are time dependent.
m 1
E total (t) = E kinetic (t) + E potential (t) = v(t)2 + m K x(t)2 . (8.17)
2 2
Yet, given absence of lossy effects and zero exchange of energy with the outside, the
total energy must remain unchanged. That is
dE total (t)
= 0 . (8.18)
dt
Using (8.17) for E total (t), (8.18) gives
dv(t) dx(t) dv(t)

m v(t) +m K x(t) = m v(t) + K x(t) = 0 .
dt dt dt
d2 x(t)
The relationship dv(t)
dt
= dt 2
leads to the relevant differential equation.
d2 x(t)
+ K x(t) = 0 . (8.19)
dt 2
8.1.4 Anharmonic Damped Motion
Assume the point mass being described is actually quite rough and its motion on the
table is subject to frictional force that is proportional to its momentum m v(t). No
external force is applied, and no other forces affect the motion of the long massless
coil. Then, the Newtonian equation of motion of the mass is
d2 x(t)
m = − Friction − Hooke’s Restoring Force
dt 2
= −(2 α) m v(t) − (m K ) x(t) , (8.20)
where (2 α) is a constant that describes the strength of friction. Dividing by m and

using v(t) = dx(t)
dt
leads to the differential equation of damped harmonic motion that
is not subject to any external force.
d2 x(t) dx(t)
2
+2α + K x(t)
dt dt
≡ D 2 + 2 α D + K x(t) = 0 . (8.21)
Solution
As is the case for all homogeneous linear (ODE), particular integral, I pi , for (8.21)
is vanishing. Given that (8.21) has constant coefficients, its characteristic equation,
E ch , can be written simply by substituting k for D.
k2 + 2 α k + K = 0 . (8.22)
As expected, for a second-order differential equation the E ch is quadratic in the

variable k. Its two roots

k = − α + α2 − K ≡ − k01 ,

k = − α − α2 − K ≡ − k02 , (8.23)
provide the complementary solution Scomp (t).
Scomp (t) → x(t) = σ1 exp (−k01 t) + σ2 exp (−k02 t) . (8.24)
Note, x(t) signifies the location of the mass at time t.

Henceforth, for convenience, we shall work only with t ≥ 0.
8.1.5 Over-Damped Anharmonic Motion
When
α2 > K (8.25)
friction outweighs the Hooke’s attraction, and motion is over-damped. Because α is

proportional to resistance to motion, α must be > 0. And because Hooke’s restoring
√therefore K , is known to be positive, for over-damped motion the inequal-
force, and
ity α > α2 − K holds. Indeed, according to (8.23), for over-damped motion the
following relationships are obtained.
k01 > 0 ; √ k02 > 0 ;

(8.26)
(k02 − k01 ) = 2 α2 − K > 0 .
Meaning, for over-damped motion both k01 and k02 are necessarily positive, and
(k02 − k01 ) is greater than zero. Therefore, irrespective of whether the constants σ1
and/or σ2 are positive or negative in (8.24), the position x(t → ∞) must tend to zero
which is the (EP). But on the other hand, at time t = 0, we have
x(0) = σ1 + σ2 . (8.27)
Should it happen that x(0) is a positive distance away from (EP), then there is a
possibility that the point mass would start moving leftward, come to a momentary
stop, and reverse direction to head back. But whether it oscillates further, or indeed
stays put on the original side never crossing over to the other, depends—see below—
on the sign and the relative size of the two constants σ1 and σ2 . As usual, two boundary
conditions are needed to fix the two constants σ1 and σ2 .
8.1.6 Both σ1 and σ2 Positive Mass Stays on Original Side
For over-damped motion, according to (8.26), if both σ1 and σ2 are positive the mass
will stay on the positive side of the (EP) and never cross over to the other side. Also,
with the passage of time the mass will continue monotonically approaching the (EP).
What about its velocity vt ?
dx
vt = = −σ1 k01 exp (−k01 t) − σ2 k02 exp (−k02 t) . (8.28)
dt
As shown in Fig. 8.1, for positive values of the constants σ1 and σ2 , at t = 0 the mass
is moving leftward toward the (EP). Its velocity approaches zero as t → ∞, and in
this process, the velocity undergoes no extrema.
Confirmation of above statement is given below. Rewrite (8.24).
vt has no extremum: xt stays

1 xt
0.5
xt and vt 0
-0.5 vt
-1
0 2 4 6 8 10 12 14
Fig. 8.1 Mass tied to an over-damped spring, with positive values of σ’s, stays on the original side

σ2
x(t) = σ1 exp(−k01 t) 1 + exp[−(k02 − k01 ) t] . (8.29)
σ1
Given σ1 is a constant, be it positive or negative, the term σ1 exp(−k01 t) does

not change sign with the passage of time. If there is to be a change in sign of the
location
x(t), it will have to come from the second term, namely
1 + σ1 exp[−(k02 − k01 ) t] . At t = 0, this term is equal to 1 + σσ21 . And as
σ2
t → ∞, because (k02 − k01 ) > 0, it tends to +1. These two numbers have the same
sign only if

σ2
1+ >0 . (8.30)
σ1
As long as

σ2
> −1 (8.31)
σ1
equation (8.30) is satisfied, and the mass stays on the original side of the (EP).
But what about its velocity? And additionally, does the velocity vt and possibly
also the location x(t) undergo any extrema?
dxExtrema
in position x(t)—if any—would occur at time t = tx:extreme such that
= 0. On the other hand, the velocity extremum—if any—would occur
dt t=tx:extreme
2
at t = tv:extreme when dv dt t=tv:extreme
= ddt x2 = 0. In other words, tx:extreme and
t=v:extreme
tv:extreme would obey the following relationships:

dx
= −σ1 k01 exp(−k01 tx:extreme ) − σ2 k02 exp(−k02 tx:extreme )
dt tx:extreme
= −σ1 k01 exp(−k01 tx:extreme )

1 + σσ21 kk02
01
exp[−(k 02 − k 01 ) t x:extreme ] =0. (8.32)

d2 x
= σ1 (k01 )2 exp(−k01 tv:extreme ) + σ2 (k02 )2 exp(−k02 tv:extreme )
dt 2 t=tv:extreme
= σ1 (k01 )2 exp(−k01 tv:extreme ))
2
σ k
1 + σ2 k022 exp[−(k02 − k01 ) tv:extreme ] = 0 . (8.33)
1 01
Items within {...} in (8.32) and (8.33) add to zero. Therefore

1 σ2 k02
tx:extreme = log − ;
k02 − k01 σ1 k01

1 σ2 k02 2
tv:extreme = log − . (8.34)
k02 − k01 σ1 k01
Since k01 and k02 are > 0, if σ1 and σ2 have the same sign, tx:extreme and tv:extreme
will not exist because logarithm of negative quantity is a complex number while
the time must necessarily be real. Therefore, under these circumstances x(t)
and v(t) will not have extrema.

σ2
8.1.7 Over-Damped Anharmonic Motion σ1 Negative But
> −1 Mass Stays on Original Side
Mass stays on the original side of the origin if (8.30) is satisfied. And this can
happen—as demonstrated in Fig. 8.2—even when σσ21 is negative as long as it

is > −1. Also because the ratio − σσ21 kk0201 is then positive, and logarithm of a
positive quantity is real, according to (8.34) extrema can occur both for the position
x and the velocity v.
vt has extremum: xt stays

3
2
xt
1
0
xt and vt
-1
-2 vt
-3
0 2 4 6 8 10 12 14
t
Fig. 8.2 An over-damped spring with positive σ1 and not too negative σ2 . Mass stays on the original
side. At time t, position xt and (ten times) velocity vt are displayed as function of t. Dots locate
extrema
8.1.8 Over-Damped Anharmonic Motion Equation (8.31)

Not-Satisfied Mass Crosses over
Next, consider systems where (8.30) is not satisfied: Meaning the requirement for
staying on the original side is violated. One class of such systems are over-damped
springs with negative σ1 and larger positive σ2 . For such systems as t increases from
zero the mass moves leftward, stops momentarily, and restarts moving in positive
direction to its final position of rest. According to (8.34), extrema in both the position
and the velocity are possible—positions indicated by dots. This behavior is portrayed
in Fig. 8.3.
8.1.9 Critically Damped Anharmonic Motion
Unlike the case where friction overpowers the Hooke’s attraction—for instance,
(8.25)—if the force due to Hooke’s attraction should happen to be of similar strength
to friction, the system would be critically damped. More precisely, this is the case
mass and velocity cross over

3
2
1
vt
0
xt and vt -1
-2 xt
-3
-4
0 2 4 6 8 10 12 14
t
Fig. 8.3 Represented here is the behavior of an over-damped spring, where (8.31) is not satisfied.
For σ1 = −10, k01 = 41 , and σ2 = 13, k02 = 21 , x(t) ≡ xt and v(t) ≡ vt are displayed as function
of t. Dots locate extrema
when K = α2 . Then, the characteristic equation, (8.22),
k 2 + 2 α k + α2 = (k + α)2 = 0 ,
has two roots,
k = −α , (8.35)
that are equal. According to well-established procedure, the solution to this differ-
ential equation can then be expressed as—[see (3.37)]—
xt = (σ3 + σ4 t) exp(k t) = (σ3 + σ4 t) exp(−α t) . (8.36)
As such
dxt
vt = = (σ4 − α σ3 − α σ4 t) exp (−α t) . (8.37)
dt
Note a dot indicates the location of the extremum (Fig. 8.4).
Critically Damped
3
2
xt
1
xt and vt 0
-1
-2 vt
0 0.5 1 1.5 2 2.5 3 3.5
t
Fig. 8.4 xt and vt of a mass being pulled by critically damped spring with σ1 = 3 and α = 2, are
displayed as function of t. Because α > 0, xt remains positive, mass stays on the original positive
side. Dots indicate extrema
8.1.10 An Exercise
Assume, much like Fig. 8.2 at time t = 0, the spring is extended and the mass is
at position x0 and has just begun moving leftward: meaning v0 = −0. According to
(8.36) and (8.37), one has
v0 = − 0 = (σ4 − α σ3 ) ,
σ4 = α σ3 ,
xt = σ3 (1 + α t) exp(−α t) = x0 (1 + α t) exp(−α t);
dxt
vt = = − x0 α2 t exp(−α t) ;
dt
dvt d2 x t
= = − x0 α2 exp(−α t) [1 − α t] . (8.38)
dt dt 2
According to (8.38), xt remains positive because α > 0. Meaning, mass stays on the
original positive side. And as t increases, mass approaches exponentially the (EP).
At t = 0, the first derivative of xt is vanishing while its second derivative is negative.
Thus, x0 is a maximum. The velocity starts at zero, moves leftwards, and as t → ∞

it again → 0. In between, the velocity reaches an extremum − x0Eα at time t = α1 .
8.1.11 Under-Damped Spring
Sinusoidal Motion
When Hooke’s attractive force is stronger than friction, i.e., K > α2 , the spring is
under-damped. The two roots of the characteristic equation are complex, namely

− k01 = − α − i K − α2 ; − k02 = − α + i K − α2 . (8.39)
The complementary solution can readily be found by using a procedure similar to

that in (3.21)–(3.24).
Scomp → xt = σ1 exp (−k01 t) + σ2 exp (−k02 t) ,

= exp (−α t) σ3 sin t K − α2 + σ4 cos t K − α2 ,

= σ5 exp (−α t) cos t K − α2 − φ , (8.40)

where σ3 = −i(σ1 − σ2 ), σ4 = (σ1 + σ2 ), and σ5 = σ32 + σ42 . Also, sin φ =

σ3
σ5
and cos φ = σσ45 .
The sinusoidal time dependence causes the mass to oscillate back and forth across
the origin. And it does so, in principle, an infinite number of times. However, because
the size of the oscillations decreases exponentially, the oscillations become all but
invisible—depending on how large α is—long before their theoretical end.
8.2 Oscillating Pendulum
Oscillatory motion is central to understanding inter-particle interaction in many phys-

ical systems. Furthermore, the differential equations needed for analyzing damped
externally driven oscillating pendula are prototypical of those used in studies of
electromagnetism, acoustics, mechanics, chemical and biological sciences, and
engineering.
We have already treated briefly the oscillatory motion of a spring on a table. More
informative is the motion of a pendulum—see Fig. 8.5—consisting of a (point-sized)
bob of mass m, that is, tied to the end of a massless stiff rod of length l. The rod
hangs down—which is the negative z-direction—from a hook that has been nailed to
Fig. 8.5 Pendulum

oscillates in two-dimensional
motion in x − z plane. When
oscillating it experiences air
resistance. At time t, it is
seen moving past angle θ θ
with the vertical
tan
θ θ
mg sin θ
mg cos θ
mg
the ceiling. The pendulum is set to oscillate in two-dimensional motion in the x − z

plane. Air resistance can be approximated as a frictional force proportional to the
speed with which the bob is moving—more precisely, α times its speed. The ensuing
friction slows the oscillatory motion. Assume that at time t the bob—see Fig. 8.5—is
moving past a position where the rod makes an angle θt with the vertical. Its equation
of motion is
d2 {l tan θt } d {l tan θt }
m 2
= −α − m g sin θt . (8.41)
dt dt
g is the acceleration due to gravity. For arbitrary size of the angle θt , (8.41) is
nonlinear. However, when θt is much less than a radian both tan θt and sin θt tend to
θt and the equation of motion becomes linear. [Note: π radians equals 180◦ . Thus,
one radian is = 57.2958◦ .] We can then write (8.41) as
d2 θt dθt
2
+ 2μ + ω02 θt = 0 (8.42)
dt dt
where

α g
2μ = ; ω0 = . (8.43)
m l
8.2 Oscillating Pendulum 241
The equilibrium position (EP) refers to the case where the angle θ is zero: Meaning,
the massless rod is exactly vertical.
Solution of (8.42)
Equation (8.42) is homogeneous linear ordinary differential equation (hlinODE).
Therefore, its particular integral, I pi , is vanishing. In order to determine its comple-
mentary solution, Scomp → θt , one needs first to solve its characteristic equation
2
E ch . To that end, dtd is replaced by k and dtd 2 by k 2 .
k 2 + 2μ k + ω02 = 0 . (8.44)
The two solutions, k ≡ k1 and k2 ,

k1 = − μ + (μ2 − ω02 ) ,

k2 = − μ − (μ2 − ω02 ) , (8.45)
of (8.44), lead to the complementary solution Scomp .
Scomp ≡ θt = σ1 exp(k1 t) + σ2 exp(k2 t) . (8.46)
For use later, it is convenient to also record its differentials.
dθt
≡ θt. = σ1 k1 exp(k1 t) + σ2 k2 exp(k2 t) ;
dt
d2 θt
≡ θt.. = σ1 (k1 )2 exp(k1 t) + σ2 (k2 )2 exp(k2 t) . (8.47)
dt 2
Because of convenience, and the fact that it does not at all affect any of the substance
of this work, in what follows we shall work with t ≥ 0.
8.2.1 Over-Damped Oscillating Pendulum
8.2.2 Angle θ t
When μ2 > ω02 , friction is overpowering. Then, the pendulum is said to be over-
damped—see Figs. 8.6 and 8.7. Because μ is greater than zero, for an over-damped
pendulum the following inequalities hold.

μ> (μ2 − ω02 ) ; k1 < 0 ; k2 < 0 ; k1 > k2 . (8.48)
Therefore, irrespective of whether the constants σ1 and/or σ2 are positive or negative,

according to (8.46) and (8.48) the angle θt must tend to the (EP) at long time. There
is also a possibility that before reaching the equilibrium position (EP) the bob will
move across to the other side, come to a momentary stop, reverse direction to head
back, and eventually θt → (E P) as t → ∞. But much depends—see below—on
the relative size of the two constants σ1 and σ2 . [Note: Knowledge of two boundary
conditions—for instance, the angle θ0 and the velocity at a specified time—is needed
to fix the two constants σ1 and σ2 .]
At t = 0, the bob is at an angle θ0 with respect to the vertical. According to (8.46)
θ0 = (σ1 + σ2 ) . (8.49)
For convenience, θ0 is chosen to be positive. The direction of motion of the bob at

t = 0 may, however, be positive or negative. Or indeed, the bob may be stationary at
that moment.
8.2.3 Angle θ t and Its Extrema
The following parameters have been set to be the same for all the four curves, A, B, C,
and D, displayed in Fig. 8.6.

1 5 k2
k1 = − ; k2 = − ; = 5 ; σ1 + σ2 = 3 ;
3 3 k1

−k1 − k2 k1 − k2 2
= μ = 1; = (μ2 − ω02 ) = . (8.50)
2 2 3
However, while the sum σ1 + σ3 = 3 is the same for all the four curves in Fig. 8.6,
individual values of σ1 and therefore σ2 may differ from one curve to the other.
8.2.4 Bob Stays on Original Side Curves A–C
To understand how the angle θt changes with t, one needs to study its description
in (8.46). Time moves forward. Given both k1 and k2 are negative, if σ1 and σ2 have
the same sign, (8.46) keeps its sign and the bob never crosses over to the other side.
Accordingly, for the curve marked A in Fig. 8.6, where σ1 = σ2 = 23 , the bob stays
on the positive side.
Regarding curve B, it is convenient to rewrite (8.46) as

σ2
θt = σ1 exp(k1 t) 1 + exp[−(k1 − k2 ) t] . (8.51)
σ1
Over Damped
6
θ 2 B
C
A
0
D
-2
0 2 4 6 8 10 12 14
t
Fig. 8.6 Pendulum motion experiences friction strong enough that its motion is over-damped. The
rod is turning past angle θ = 3◦ at time t = 0
Clearly, the term σ1 exp(k1 t) does not change sign with passage of time. And
θt were
if to change sign, it will have to come from the second
term, namely
1 + σσ21 exp[−(k1 − k2 ) t] . At t = 0, this term equals 1 + σσ21 . And at t → ∞,
because (k1 − k2 ) > 0, it is equal to 1. These two values can have the same sign
only if

σ2
> −1 . (8.52)
σ1
The curve marked B in Fig. 8.6 refers to a pendulum that has σ1 = 3◦ × 45 , σ2 =

−3◦ × 41 . Therefore, σσ21 = − 15 is within the range specified by (8.52). Hence,
the bob stays on the positive side. To determine whether the angle θt experiences
an extremum, it is convenient to reprint (8.34) in terms of the present notation. It
indicates the angle, and the angular velocity undergoes extremum at times

1 σ2 k 2
tθt:extremum = log − ;
k1 − k2 σ1 k 1

1 σ2 k 2 2
t dθt :extremum = log − . (8.53)
dt k1 − k2 σ1 k 1

σ2
Given (k1 − k2 ) = 43 , k2
k1
= 5, and σ1
= − 15 ,

3 1 3
tθt:extremum = log (5) = log (1) = 0 . (8.54)
4 5 4
At time zero, according to (8.46) and (8.49), the angle is 3◦ . Its second derivative,

d2 θ t 15 −1 2 3 −5 2
= σ1 (k1 )2 + σ2 (k2 )2 = −
dt 2 t=0 4 3 4 3
5
= − 4σ1 k12 = − , (8.55)
3
is negative. Meaning, the bob starts its journey at θtθt:extremum = 3◦ , that is a maximum,
and its distance from the (EP) continues decreasing monotonically with the passage
of time.
For the curve C in Fig. 8.6, σ1 = 10, σ2 = −7 and the ratio σσ21 = −0.7 is again
> −1. Thus, the bob stays on the positive side. Initially, at time tθt:extremum , the angle
continues increasing and according to (8.53) reaches a maximum θt:extremum , where
it halts momentarily and starts decreasing.

3
log [0.7 (5)] = 0.939572 .
tθt:extremum =
4
θt:extremum = σ1 exp (k1 tθt:extremum ) + σ2 exp (k2 tθt:extremum )

1 5
= 10 exp − 0.9396 − 7 exp − 0.9396 = 5.850◦ .
3 3
Eventually—literally, at t → ∞—the bob arrives (EP).

Curve D refers to the case σ1 = −6◦ , σ2 = 9◦ . As a result σσ21 = −1.5 > −1
and the bob crosses over to the other side before it stops at time, say, tmin at an angle,
say, θmin , reverses direction and moves toward the (EP) which is where its journey
ends.
According to (8.53),

3 σ2 3
tmin → = tθt:extremum = log − × 5 = log(1.5 × 5)
4 σ1 4
= 1.511 , (8.56)
and (8.46),
θmin = σ1 exp (k1 tmin ) + σ2 exp(k2 tmin ) ,

1 5
= − 6 exp − × 1.511 + 9 exp − × 1.511 = − 2.901◦ . (8.57)
3 3
8.2.5 Extremum in θ t .

The x axis in Fig. 8.7 is the time t, and the y axis is the angular velocity dθ
dt
t
≡ θt .
[See (8.47)].
Curves A, B, C, and D in Fig. 8.7 refer to the same parameters as those in Fig.
8.6. Because θt in curve A, Fig. 8.6, undergoes no extremum, curve A in Fig. 8.7 does
not pass through θt . = 0. On the other hand, because of the presence of an extremum
Over−Damped Pendulum
3
1 D
0
-1 B
A C
-2
-3
-4
0 2 4 6 8 10
Angular Velocity versus Time

Fig. 8.7 x axis is time t and y axis is angular velocity θt . . Curves A, B, C, and D pertain to the
same values as corresponding curves in Fig. 8.6. Angular velocity passes through zero where angle
θt reaches extremum
in θt in each of the other three curves in Fig. 8.6, the relevant curves pass through
θt . = 0. For curve B, this occurs at the beginning: that is, at t = 0. For curves C and
D, θt . = 0 occurs at 0.9396 and 1.511, respectively.
In addition to θt . itself, graphical treatment of extrema for θt . requires knowledge
also of θt .. —these are both available in (8.47)—as well as that of θt ... , which is given
below.
d3 θ t d2 θ t .
θt ... ≡ = = σ1 (k1 )3 exp(k1 t) + σ2 (k2 )3 exp(k2 t)
dt 3 dt 2

σ2 k23
= σ1 (k1 ) exp(k1 t) 1 +
3
exp[−(k1 − k2 ) t] . (8.58)
σ1 k13
Is there an extremum present in curve A, Fig. 8.7? To answer this query, use (8.53)
and calculate the time t = tθt:extremum
when such an extremum would occur. To that end
recall that (k1 − k2 ) = 43 , and kk21 = 5. Next set θt .. = 0 in (8.47) and rewrite the
result as

1 σ2 k22 3 σ2
tθt:extremum = log − = log − 25 . (8.59)
k1 − k2 σ1 k 1
2 4 σ1

σ2
For curve A, σ1
= 1.5
1.5
= 1. Therefore,

3
tθt:extremum = log (−25) . (8.60)
4
This is a complex number. Because time tθt:extremum has to be real this result is unphysical.
Thus, curve A does
not undergo any extrema.
For curve B, σ1 = −3/4
σ2
15/4
= −1/5. Therefore, an extremum occurs at time

3 −1 3
tθt:extremum = log − 25 = log (5) = 1.20708. (8.61)
4 5 4
Using this value of time in (8.46), one can determine the relevant value of θt . at this
extremum. The result is −0.668739. Because θt ... at this time—namely 1.20708—
d3 θ
θt ... ≡ = σ1 (k1 )3 exp(k1 t) + σ2 (k2 )3 exp(k2 t)
dt 3
= 0.371521, (8.62)
is positive, the extremum

is a minimum.
For curve C, σσ21 = −710
. Therefore, an extremum occurs also here. The relevant
time is

3
t = log (5) = 2.14665. (8.63)
4
Proceeding as before, the θt . for this extremum is −1.303795 and again it is a mini-
mum.
For curve D, the extremum is a maximum and it occurs at t = 2.18256 and
θt . = 0.64565.
8.2.6 Critically Damped Pendulum
When strength of friction becomes similar to the ordering tendency of the natural
vibrations a pendulum is said to be critically damped—see Figs. 8.8 and 8.9. More
precisely, this is the case when μ2 = ω02 . Here, the characteristic equation (8.44)
has two equal roots : k1 = k2 = −μ. [See (8.45)] Therefore, according to the well-
established procedure, and (3.37), the angle at time t can be expressed as
Critically Damped
4 A
3
2 B
θ 1
0
-1
-2 C
0 5 10 15 20
t
Fig. 8.8 Angle pendulum rod of critically damped pendulum makes it displayed as function of
time. Curves A, B, and C are defined in (8.69)–(8.74)
Angular Velocity: Critically−Damped

2
A
C
0
B
-1
0 2 4 6 8 10
t
dt , for a critically damped pendulum.
dθ
Fig. 8.9 Plotted along the vertical axis is the angular velocity,
Curves A, B, and C are the same as in Fig. 8.8
θt = (σ3 + σ4 t) exp(−μ t) . (8.64)
Prediction from this equation is displayed in Fig. 8.8. At time t = 0, the bob is at
a point where the rod, according to (8.64), makes an angle θ0 = σ3 (which was set
equal to +3◦ ). Because μ > 0, as time t → ∞, the angle θt tends to (EP). In between,
there are two possibilities: First, the angle stays positive, eventually reaching (EP).
And the second, the angle switches over to the negative side where at time, say tcritical ,
the angle reaches, say θtcritical , where it comes to a momentary stop. Instantly, the bob
starts the reverse journey, eventually reaching the end of its travel at the midpoint:
that is, when the angle θt reaches the (EP).
Let us examine this behavior and see exactly what transpires. To this end, in
Fig. 8.9, look at the derivative of θt .
dθt
= (σ4 − μσ3 − μ σ4 t) exp(−μ t) . (8.65)
dt
Because dθt
dt
must tend to zero at t = tcritical , (σ4 − μσ3 − μ σ4 tcritical ) = 0. As such

σ4 − μσ3 1 σ3
tcritical = = − . (8.66)
μ σ4 μ σ4
Therefore, according to (8.64) and (8.66),
θtcritical = (σ3 + σ4 tcritical ) exp(−μ tcritical )

σ4
= exp (−μ tcritical ) . (8.67)
μ
In order to determine the type of extremum that θtcritical represents, one needs to look
at the second derivative ddtθ2t at time t = tcritical . One knows that friction is positive
2
therefore the parameter μ > 0. And if the parameter σ4 should also be > 0, the
second differential of equation (8.65) would be clearly negative.

d2 θ t 1 σ3
= exp −μ − (−μ σ4 ) < 0, (8.68)
dt 2 t=tcritical μ σ4
As such, θtcritical would be a maximum. At this juncture, it is helpful to give θtcritical

a more informative new name: call it θmax . For the same reason, also re-name the
relevant tcritical as tmax . And rewrite (8.66) and (8.67) as

σ4 − μσ3 1 σ3
tmax = = − . (8.69)
μ σ4 μ σ4
and

σ4
θmax = exp (−μ tmax ) . (8.70)
μ
Such behavior is demonstrated in curve A in Fig. 8.8. For this curve, σ3 = σ4 = 3◦

and μ1 = 3. Therefore

1 σ3
tmax = − = 2 , (8.71)
μ σ4
and

σ4 ◦ 2
θmax = exp (−μ tmax ) = 3 × 3 × exp − = 4.621◦ . (8.72)
μ 3
Next, in Fig. 8.8, examine curve B where σ3 = 3◦ , μ1 = 3, and σ4 = 0.5◦ . But now,
because the predicted tmax is equal to −3, which is a time before the experiment
began, the requirement for an observed maximum is not satisfied. As a result, while
the angle stays positive, it continues to decrease monotonically heading to the (EP)
as t increases toward ∞.
The third curve, C in Fig. 8.8, refers to the case where = 3, σ4 = −σ3 = −3.
1
μ
2
d θtcritical
Thus, μ σ4 is < 0. With reference to (8.68), (−μ σ4 ) , and therefore dt 2 t=tcritical
are positive. As a result, the relevant θcritical is a minimum. Again, it is helpful to call
this value of θcritical by a new name: θmin , and the time it is reached as t = tmin . Thus

1 σ3 3
tmin = − = 3− = 4 , (8.73)
μ σ4 −3
and

|σ4 | 4
θmin = − exp (−μ tmin ) = − (3◦ × 3) exp − = −2.372◦ (8.74)
μ 3
With the progress of time, due to the presence of negative exponent in (8.64), θt →
(E P).
8.2.7 Under-Damped Motion
When friction is weaker than the critical amount, i.e., ω02 > μ2 , the pendulum is
under-damped—see Fig. 8.10. The two roots, k1 and k2 —compare (8.45)—of the
characteristic equation (8.44) are complex, namely

k1 = −μ − i ω02 − μ2 ; k2 = −μ + i ω02 − μ2 . (8.75)
Using a procedure similar to that in (3.21)–(3.24), the complementary solution is

readily found.
Scomp → θunder-damped = σ1 exp (k1 t) + σ2 exp (k2 t) ,

= exp (−μ t) σ3 sin t ω0 − μ + σ4 cos t ω0 − μ
2 2 2 2 ,

= σ5 exp (−μ t) cos t ω0 − μ − φ
2 2 , (8.76)

where σ3 = −i(σ1 − σ2 ), σ4 = (σ1 + σ2 ), and σ5 = σ3 + σ4 . Also, sin φ =
2 2

σ3
σ5
and cos φ = σσ45 . Because μ > 0, as t → ∞ the angle θunder-damped tends
to (EP). However, before the bob comes to the final absolute stop, it repeats the
following performance many times—in principle, infinite number of times.
At time t = 0 and angle 3◦ , consider the bob is heading left. As it moves toward
decreasing angle, it crosses the (EP), proceeds further left, slows down for a momen-
tary halt, reverses its direction, and starts moving forward to the right: crosses the
(EP) again before coming to a momentary stop on the right side and starting another
journey leftward. This behavior contrasts with that of the over-damped and critically
damped pendulums that either never move across the midpoint or cross it at most
only one time.
Another point to note is that according
to (8.76),the angle θunder-damped reaches a
maximum = θn at a time t = tn if tn ω02 − μ2 − φ = 2nπ and the next maximum

= θn+1 at time tn+1 when tn+1 ω0 − μ − φ = 2(n + 1)π. Because the cosines
2 2
at both these times are equal to unity, the ratio of the angles at these two successive
maxima is
⎛ ⎞
θn exp (−μ tn ) 2πμ
= = exp ⎝ ⎠ . (8.77)
θn+1 exp (−μ tn+1 ) ω 2 − μ2 0
[Note : The same is also true for the ratio of angles at two successive minima.]
Remarkably, this ratio is not dependent on n. Rather, it is the same for any two
successive maxima (or even, any two successive minima). Also, it is readily measur-
able. Another quantity that is easy to measure is the cycle time, δcycle , of the damped
oscillation. That is
⎛ ⎞
2π
δcycle = tn+1 − tn = ⎝ ⎠ . (8.78)
ω02 − μ2

θn
From δcycle and the ratio log θn+1
, one can determine the strength of friction,
⎛ ⎞
θn
log θn+1
μ = ⎝ ⎠, (8.79)
δcycle
which is otherwise hard to measure.

For an under-damped pendulum, a typical plot of the angle versus time could be
that given in Fig. 8.8. Here, φ has been set = 0, and σ5 set at 3◦ . Also, μ1 is chosen

10◦
to be 1.5 and ω02 − μ2 is set at time . In other words, Fig. 8.8 refers to

θud:transient = σ5 exp (−μ t) cos t ω0 − μ
2 2

t
= 3 exp − cos(10 t) . (8.80)
1.5
Fig. 8.10 With sinusoidal external force, angle θ the pendulum makes with the vertical is displayed
as a function of time. Plotted here is (8.80)
Therefore, according to (8.77), in Fig. 8.10 the ratio of any

two successive maxima,
or minima, for the angle θud:transient is equal to exp 1.5×10
2π
= 1.520.
8.2.8 Steady-State Motion
Transient state motion of damped pendulums was analyzed in the foregoing. The
steady-state motion, caused by the application of an applied force, is considered
below.
In the presence of an applied force, say m l f (t), the original equation of motion,
namely (8.41), changes to the following:
d2 {l tan θ} d {l tan θ}
m 2
+α + m g sin θ = m l f (t) . (8.81)
dt dt
For small angle θ, it can be expressed as
d2 θ dθ
2
+ 2μ + ω02 θ = f (t) (8.82)
dt dt

g α
where ω0 = l
and μ is = 2m
.
8.2.9 Sinusoidal External Force: Steady State
Oscillating Pendulum
Assume the externally applied force is sinusoidal in time with angular velocity ω.
m l f (t) = m l A cos(ω t) ; i.e. ; f (t) = A cos(ω t) . (8.83)
With the application of external force (8.83), the equation of motion (8.82) becomes
inhomogeneous linear (ODE) with constant coefficients. The external force, how-
ever, does not affect the complementary solution that relates to the transient states
discussed in detail in the foregoing. Therefore, only the particular integral, I pi , that
leads to steady-state motion needs to be evaluated here. Denoting = dtd , the par-
ticular integral of equation (8.82) is calculated in the usual fashion as follows.
1
θ pi → I pi = A [cos(ω t)]
2 + 2μ + ω02
1/2
= A [exp(i ω t) + exp(−i ω t)]
2 + 2μ + ω02
1/2 1/2
= A exp(i ω t) + exp(−i ω t)
−ω 2 + 2μ i ω + ω02 −ω 2 − 2μ i ω + ω02
(ω02 − ω 2 ) cos(ωt) + 2μω sin(ωt)
= A
(ω02 − ω 2 )2 + 4μ2 ω 2

A cos(ωt − )
A cos(ωt − ) ω02
= =
(ω02 − ω 2 )2 + 4μ2 ω 2 2 2 2 2
1 − ωω0 + 2μ
ω0
ω
ω0
= Mratio × θstatic deflection × cos(ωt − ). (8.84)

The following expression in (8.84) is the magnification ratio Mratio .
1
Mratio =
2 2 2 2
ω ω
1− ω0
+ 2μ
ω0 ω0
1
= , (8.85)
4 2 2
ω ω 2μ
1+ ω0
− ω0
2− ω0
The magnification ratio is unity when the applied force is constant in time: that is,
Mratio = 1 when ω = 0. And when the frequency of the applied force reaches the
undamped natural frequency of the pendulum, that is when ω → ω0 , Mratio → 2ωμ0 .

Next, in (8.84), is the quantity ωA2 which is the static deflection, θstatic deflection . This
0
nomenclature is owed to the fact that in the absence of time-dependent motion the
applied force would deflect the equilibrium position of the bob through an angle
≈ mmgA = ωA2 . Also, in (8.84), is the so-called lag angle defined by the relations
0
(ω02 − ω 2 )
cos() = ,
(ω02 − ω 2 )2 + 4μ2 ω 2
2μω
sin() = . (8.86)
(ω02 − ω 2 )2 + 4μ2 ω 2
For convenience, let us introduce the notation

2 2
ω 2μ
x= ; σc = 2− ;
ω0 ω0
z = 1 + x 2 − x σc . (8.87)

d2 z
Given the second derivative of z is a positive constant for all x, namely dx 2
= 2,
find the location where the first derivative of z is equal to zero. That is

dz
=0
dx x=xc
= (2xc − σc ) . (8.88)
This defines a location xc ,

2 2
σc 1 2μ μ
xc = = 2− = 1−2 , (8.89)
2 2 ω0 ω0
where z reaches a minimum, and as a result, Mratio reaches a maximum.
1 1 1
(Mratio )max = √ = = . (8.90)
z 1− σc 2
1 − xc2
4
Note when friction decreases, xc increases toward unity. As a result, 1 − xc2 decreases
and the magnification ratio increases. Indeed, when there is no friction at all present,
meaning when μ = 0, (Mratio )max → ∞.
Chapter 9
Resistors, Inductors, Capacitors
Introduced first are Kirchhoff’s two rules that state: ‘The incoming current at any
given point equals the outgoing current at that point,’ and ‘The algebraic sum of
changes in potential encountered by charges traveling, in whatever manner, through
a closed-loop circuit is zero.’ Included next is the Ohm’s law: ‘In a closed-loop
circuit that contains a battery operating at V volt, and a resistor of strength R ohms,
current flow is I amperes: I = VR .’ Several problems relating to additions of finite
numbers of resistors, placed in various configurations, some in series and some in
parallel formats, are worked out—see (9.2)–(9.30) and Figs. 9.1, 9.2, 9.3, 9.4, 9.5
and 9.6. More involved problems relating to total resistance and current flows in
infinite networks of resistors are treated next—see (9.31)–(9.60) and Figs. 9.7, 9.8,
9.9, 9.10, 9.11, 9.12, 9.13, 9.14, 9.15, 9.16, 9.17. Electric circuit elements are listed
in (9.61). Inductors are introduced in (9.62). Series and parallel circuits constituted
of finite number of inductors, and infinite series–parallel circuits of inductors are
treated in (9.63)–(9.70) and displayed in Figs. 9.18 and 9.19. The same is done
also for capacitors—see (9.71)–(9.82) and Figs. 9.20 and 9.21. Resistor–capacitor
circuits are treated in (9.83) to (9.86), and some results are displayed in Fig. 9.22.
Resistor–inductor circuits are studied in (9.91)–(9.96), and results are plotted in
Fig. 9.23. Inductor–capacitor circuits are analyzed in (9.97)–(9.106), and results are
displayed in Fig. 9.23.
9.1 Electric Current
9.1.1 Kirchhoff’s Rules
Central to the understanding of current flow are a few rules that are obeyed by the
current and the circuitry through which it flows.

https://doi.org/10.1007/978-3-319-76406-1_9
258 9 Resistors, Inductors, Capacitors
Fig. 9.1 With the switch on,

the battery operates at
constant voltage V and
drives current I amps through
resistors R3 , R2 , R1 that are
connected in series
Fig. 9.2 Battery operating at (b)

voltage V drives currents
I1 , I2 , . . . , In amps through
resistors R1 , R2 , . . . Rn
connected in parallel
Fig. 9.3 Shown is the dd, (c)

the duodectet, that in
addition to twelve resistors
also has a switch and a
battery
Fig. 9.4 Shown is circuit, an (d)

on–off switch, a battery
operating at V volts, and
variously put together four
resistors: R1 , R2 , R3 and R4
9.1 Electric Current 259
Fig. 9.5 This replicates the (e)

circuit in Fig. 9.4. With the
switch on, the battery drives
current I1 through resistors
(R1 + R22 ) and I2 through
( R44 + R33 )
Fig. 9.6 Points W and Z are (f)

a single point WZ. Effective
resistances between X and
WZ and between WZ and Y
are Rtotal X −W Z and
RtotalW Z −Y
Fig. 9.7 For convenience,

current is drawn as I

current is drawn as I
Kirchhoff’s First Rule

(a): Electric charges, like billiard balls, flow down from points at higher electric
potential to points at lower potential.
(b): When positive electric charge is placed at a given point, electric potential at
the point rises. And the charge stays at the point only if the neighborhood is at even
higher potential or if the charge is placed in a capacitor that is designed for the task

current is drawn as I. With
the switch on, the battery
drives current I through the
infinite network shown. At
the end of its journey, current
I returns

current is drawn as I. Figure
9.10 circuit represents the
essentials of Fig. 9.9. The
resistance Reffect0 equals the
resistance of the infinite
network that would remain
after the first-trio, R, R0 , R
has been subtracted

the current is drawn as I.
With the switch on, the
battery operating at voltage
V drives current I through
the infinite array of resistors
shown

the current is drawn as I.
With the switch on, the
battery drives current I
through the infinite array of
resistors shown in Fig. 9.11.
The effective resistance of
the infinite circuit is RNORC
Fig. 9.13 With the switch

on, the battery drives current
through an infinite number of
resistors R1 and R2 arrayed
in the manner displayed

the current is drawn as I. In
the ‘septet’ the current I
splits into three parts: I1 , I2 ,
and I − I1 − I2

the current is drawn as I. In
the ‘equivalent-sextet,’ the
battery, operating at voltage
V, drives current I that splits
into I1 and I − I1

current is drawn as I. In the
‘equivalent-triplet,’ the
through resistors R1 , Rq and
R1

current is drawn as I. In the
‘equivalent-singlet,’ the
through resistor
Re f f ect − Cc which is the
effective resistance of the
infinite circuit shown in
Fig. 9.13
Fig. 9.18 An infinite array

of three-upon-three
inductors, L 1 , L 2 , L 3 , is put
together in the manner shown
Fig. 9.19 The effective

inductance of the infinite
array of inductors shown in
the Fig. 9.18 is equivalently
represented by inductance of
a single inductor L eff
of holding charge. Lacking the capacitor, electric charge does not accumulate at a
point. As such, at any point, the inflow of charge during a given time interval equals
the outflow of charge during the same time interval.
The rate of flow of charge, (q + δq − q), during a time interval (t + δt − t)—that
is, δq
δt
—is called electric current at the given time t. Thus, except for the extraordinary
circumstance referred to above, incoming current at any point equals the outgoing
current at that point. This is called Kirchhoff’s first rule.
Fig. 9.20 An infinite array

of three-upon-three
capacitors, C1 , C2 , C3 , is
put together in the manner
displayed
Fig. 9.21 Effective

capacitance of the
three-upon-three capacitors
C1 , C2 , C3 , drawn in
Fig. 9.20 is Ceff
Fig. 9.22 Displayed is a

circuit connected in series to
an on–off switch, a battery, a
resistor R, and a capacitor C.
At time t the current is i(t)
and the charges on the plates
are q(t) and −q(t)
Kirchhoff’s Second Rule

Consider a person who starts off walking at a given point and engages in closed-loop
travel: Meaning, he walks for however long, in whatever manner, up and down and
sideways, but arranges to return exactly to the same point where he started. Adding
diligently the changes in his height during the closed-loop travel, the sum must surely
be zero because the person ends up exactly at the same height as he started.
Fig. 9.23 Displayed is a

circuit with a battery that at
time t supplies current i(t) to
a series connection
constituted of an inductor L,
a capacitor C, and an on–off
switch
Kirchhoff’s second rule asserts that much like this person, if a given charge travels
in a closed-loop circuit in whatever manner, the algebraic sum of changes in potential
encountered must be zero.
9.1.2 Ohm’s Law
According to Ohm’s law, current I amperes flows through a closed-loop circuit that
consists of a battery operating at voltage V volts and a resistor of strength R ohms.
V volts
I amper es = . (9.1)
R ohms
9.1.3 Addition of Resistors
Three Resistors in Series

With the switch on, the large battery, shown in Fig. 9.1, operates at V volts and
drives current I amps through three resistors of strength R1 , R2 , and R3 ohms that
are connected in series, thereby forming a closed-loop series circuit. The physical
result of the current flow through the circuit is found not to depend upon which of
the three resistors leads, which is in the middle, or which follows. To understand
the behavior of this series circuitry, consider a person trekking through a tunnel
of length d1 and immediately thereafter through another similar tunnel of length
d2 and yet another third similar tunnel of length d3. Clearly, it would take the
person same amount of effort as trekking through a similar single tunnel of length
(d1 + d2 + d3). Accordingly, the effective resistance of three resistors, R3 , R2 , and
R1 ohms placed one after the other—that is, as shown in Fig. 9.1, in series—is
Rseries−3 (e f f ective) = (R3 + R2 + R1 ) ohms.
n Resistors in Series
By an argument similar to that given above, the effective resistance of n resistors
R1 , R2 , . . . , Rn connected in series should be R1 + R2 + · · · + Rn . However, more
formally, we can argue as follows. According to ohm’s law23. , when a current I
amper es 31. flows through a resistor of strength R1 ohms the potential across it
decreases I R1 volts 32. . When current I flows through n resistors in series, the
potential decrease across R1 is I R1 , that across R2 is I R2 , . . . , and so on. There-
fore, the total decrease in potential across the n resistors is I (R1 + R2 + · · · + Rn ) :
exactly the same as it would be across a single resistor Rseries−n (e f f ective).
Rseries−n (e f f ective) = R1 + R2 + · · · + Rn . (9.2)
For large number of resistors in series, the effective total resistance is large.
2 Resistors in Parallel
Consider two resistors R1 and R2 joined together in parallel and connected to a
battery supplying direct current at constant voltage V volts. The parallel connection
implies that while the positive outlet of the battery is connected to one end of each
of the two resistors, the negative outlet is connected to the other end of the same
two resistors. As a result, the potential difference across each of the two resistors is
identical: equal to V volts. Assume this process results in current I1 across resistor
R1 and I2 across R2 . Then according to Ohm’s law
V
I1 =
R1
V
I2 = . (9.3)
R2
The total current, I parllel−2 (e f f ective), that flows through the two resistors R1 and
R2 that are connected in parallel, and their total resistance,
R parallel−2 (e f f ective), are

V V
I parllel−2 (e f f ective) = I1 + I2 = +
R1 R2
V
≡ . (9.4)
R parallel−2 (e f f ective)
Eliminating V from (9.4) leads to the relationship
1 1 1
= + . (9.5)
R parallel−2 (e f f ective) R1 R2
The resistance of two resistors R1 and R2 connected in parallel, namely
R1 . R2
R parallel−2 (e f f ective) = R1 R2 ≡ (9.6)
R1 + R2
has some interesting features. [Note: The notation a b stands for a+b
a.b
.] To examine
these features, keep R1 constant and make changes in R2 .
Process (1) : Consider the case RR21 1. To this end, re-write (9.6) as
R2
R parallel−2 (e f f ective) = (9.7)
1 + RR21

and expand it in powers of the small parameter RR21 .
2
R2 R2
R parallel−2 (e f f ective) = R2 1 − +O . (9.8)
R1 R1
This result is interesting. When two resistors are connected in parallel, and one of the
resistors is much smaller than the other one, the effective resistance is even smaller
than the smaller resistor !
Process (2) : Increase R2 till R2 = R1 . Even though there are two resistors of
strength R1 , their total resistance is only half of R1 .

R1
R parallel−2 (e f f ective) = . (9.9)
2 lim R2 →R1
Process (3) : Not surprisingly, an increase in R2 beyond R1 increases

R parallel−2 (e f f ective) beyond R21 . But, the rate of increase is even less than a

quarter of what one might expect. For instance, for a small increase R2R−R 2
1
1:
⎡ ⎤

R1 ⎣ 1
R parallel−2 (e f f ective) = R1 R2 = ⎦
2 R2 −R1
1− 2 R2
2
1 R2 − R1 R2 − R1
= R1 + −O . (9.10)
2 4 R2 2 R2
Process (4) : With further increase in R2 , the effective resistance, albeit slowly,
rises further. When R2 R1 , (9.6) gives
2
R1 R1 R1
R parallel−2 (e f f ective) = = R1 1− +O . (9.11)
1 + RR21 R2 R2
The effective resistance reaches R1 as R2 → ∞.

n Resistors in Parallel
Current flowing through n resistors R1 , R2 , . . . , Rn that are joined together in par-
allel is not unlike very large number of people wanting to tread across a mountain
through a group of n tunnels that begin and end in very close proximity. And if
the tunnels are of different widths, surely most people would choose to go through
the tunnel that is the widest. And similar choices would be made for other tunnels.
Same must also be the case for current flowing through n resistors in parallel so that
most current would flow through that resistor that has the smallest resistance, and
so on. Assuming the potential difference is V volts across the parallel assembly—
which means the potential difference is V volts across each of the n resistors in the
assembly—and currents I1 , I2 , . . . , In flow through the n resistors that are linked in
parallel, then according to Ohm’s law
V = I1 R1 = I2 R2 = · · · = In Rn . (9.12)
The above can be represented more compactly as
V
Ii = , i = 1, 2, . . . , n .
Ri
As a result the total current, Iparallel−n (e f f ective), the effective total resistance,
Rparallel−n (e f f ective), and the potential V are related as follows:
V V V V
I1 + I2 + · · · + In = + + + ··· +
R1 R2 R3 Rn
V
≡ Iparallel−n (e f f ective) ≡ . (9.13)
Rparallel−n (e f f ective)
Thus
1 1 1 1
= + + ··· + . (9.14)
Rparallel−n (e f f ective) R1 R2 Rn
This is an interesting result. To a layman, some of its consequences would be

counter-intuitive. For instance, if all the n resistors were identical, each having resis-
tance R, then according to (9.14)
1 n
= ,
Rparallel−n (e f f ective) R
or equivalently
R
Rparallel−n (e f f ective) = . (9.15)
n
In other words, for very large number of resistors, all in parallel, the effective total
resistance is very small.
A Few Resistors Together
Process (1) : A group of three resistors R1 , R2 , R3 ohms—to be called the ‘origi-
nal trio’—is connected in series across an on–off switch and a battery operating at
constant voltage V volts—see Fig. 9.3. Turning the switch on makes it a closed-loop
circuit.
Process (2) : Three additional resistors of strength R1 ohms each are connected
in parallel to the resistor R1 of the original trio : thus forming a sextet.
Process (3) : The sextet is enhanced further to form a nonet by connecting in
parallel, to the resistor R2 of the original trio, three new resistors of strength R2 each.
Process (4) : In a fashion similar to the above, another three resistors of strength
R3 ohms each are added next. Each of these new resistors R3 is connected in parallel
to the resistor R3 of the original trio. This process results in forming a duodectet to
be called dd, composed of a total of twelve resistors, a switch, and a battery.
Problem (A)
Calculate the total effective resistance, Reffect−dd , of the duodectet dd demonstrated
in Fig. 9.3.
9.1.4 Solution (A)
The four resistors R1 in parallel can equivalently be replicated by a single resistor

Reffect−dd1 .
R1
Reffect−dd1 = . (9.16)
4
Similarly, the four resistors R2 in parallel are replaced by a single resistor Reffect−dd2 .
R2
4
And finally, the four resistors R3 in parallel can be replaced by a single resistor
Reffect−dd3 .
R3
4
In the manner described above, the dd may be replaced by a new triplet. The
effective resistance of such triplet would be Reffect−dd .
Reffect−dd = Reffect−dd1 + Reffect−dd2 + Reffect−dd3

R1 + R2 + R3
= . (9.19)
4
Figure 9.4 describes a circuit with an on–off switch, a battery operating at V volts,
and four resistors: R1 , R2 , R3 and R4 .
Left part of the circuit refers to points W, X, and Z. There is a single resistor of
strength R1 between points W and X. Connected in series to R1 are two resistors of
strength R2 ohms. These two resistors are set in parallel between points X and Z .
As a result, the left part of the circuit—that is, W → X → Z —can equivalently be
expressed as having only two resistors, R1 and R22 , connected in series.
The part of the circuit on the right—that is, between the points W, Y, and Z—has
seven resistors. Four are of strength R4 ohms each, and they are connected in parallel
to each other. The other three are of strength R3 ohms each, and they too are joined
together in parallel to each other. Accordingly, this part of the circuit can also be
expressed as having only two resistors that are connected in series. Their strengths
are R44 , from W → Y, and R33 , from Y → Z .
In the manner described above, Fig. 9.5 effectively replicates Fig. 9.4. When the
switch is on, the battery—operating at constant voltage V volts—drives current I1
amps through resistors R1 and R22 , and I2 amps through resistors R44 and R33 .
9.1.5 Problem (B)
In the circuit shown in Fig. 9.5, turn the switch on and work out currents I1 and I2 , the
potential difference VX − VY , and the effective resistance, Reffect−XY between points
X and Y. The relevant data is as follows:
V = 10 volts ;
R1 = 15 × 103 ohms ; R2 = 70 × 103 ohms ;
R3 = 60 × 103 ohms ; R4 = 20 × 103 ohms . (9.20)
V is the battery voltage, and R1 → R4 are the resistors described above. Their
strength is as specified in (9.20).
9.1.6 Solution (B)
Shown in Fig. 9.5 is current flow, = I amps, from point W to point Z .
I = I1 + I2 , (9.21)
V V
I1 = , I2 = . (9.22)
R1 + R2
2
R3
3
+ R4
4
The potentials at points X and Y are

R2
V X = I1 ,
2

R3
VY = I2 . (9.23)
3

[Note : Prove VX = V − R1 I1 and VY = V − R44 I2 also lead to the same result
as (9.24).] The effective potential difference, Veffect−XY , between X and Y is
V R2 V R3
Veffect−XY = VX − VY = − . (9.24)
2 R1 + R2 R3 + 3 4R4
Following (9.24) and (9.22), and using the numbers provided in (9.20), leads to
Veffect−XY = (0.7 − 0.8) volt = − 1 volt ;

I1 = 0.2 × 10−3 amps; I2 = 0.4 × 10−3 amps. (9.25)
Knowing the effective potential difference, Veffect−XY, and the currents I1 and I2 , the
effective resistance, Reffect−XY, between points X and Y can now be calculated. The
effective current flow from X to Y is Ieffect−XY.
Ieffect−XY = I1 − I2 = − (0.2) × 10−3 amps . (9.26)
Therefore
Veffect−XY
Reffect−XY = = 5 × 103 ohms . (9.27)
Ie f f ect−X Y
9.1.7 Problem (C)
Figure 9.6 shows a circuit that represents a slight change from the circuit in Fig. 9.5.
Here the battery is missing, the on–off switch is absent, and the labels i 1 and i 2
have been removed. Thereby, in effect, connecting directly the points W and Z. [This
effective double point is called the point WZ.] Calculate effective resistance between
X and WZ and between WZ and Y.
9.1.8 Solution (C)
Detailed behavior of this ‘slightly changed’ circuit—shown in Fig. 9.6—is substan-

tially different from that of its parent circuit 1e. Here the point W is in fact coincident
with point Z, making the two a single point WZ.
Each of the couplets (R1 + R22 ) and ( R33 + R44 ) are now in parallel. Accordingly,
the resistances between points X and WZ—that is Rtotal X −W Z —and that between
WZ and Y, are
Rtotal X −W Z = R1 (R2 /2) , (9.28)
and
RtotalW Z −Y = (R4 /4) (R3 /3) . (9.29)
The data provided in (9.20) yields
21
Rtotal X −W Z = × 103 ohms .
2
RtotalW Z −Y = 4 × 103 ohms . (9.30)
9.2 Infinite Networks of Resistors
Resistance and Current

Work out the effective total resistance and currents in infinite assemblies of resistors
described below. Consider first the circuit shown in Fig. 9.7.
Figure 9.7 is an infinite network of resistors, each of strength R ohms, connected
together as follows.
Process (1) : Three resistors R, to be called the first-trio, are connected in series
across a large battery operating at V volts and an on–off switch. Turning the switch
on creates a closed-loop circuit.
Process (2) : The middle resistor R of the first-trio is connected in parallel to a
similar second-trio.
Process (3) to Process (∞) : Process (2) is repeated ad infinitum.
When the switch is turned on, the battery notices only a single effective resistance
Reffective ohms. Reffective is the sum of two resistors in series—each of strength R
ohms—and two resistors, R and Reffective , in parallel. Thus
Reffective = 2R + R Reffective . (9.31)
Equation (9.31) leads to a quadratic,

2
Reffective = 2R ∗ Reffective + 2R 2 , (9.32)
which has one positive solution

√
Reffective = R 1 + 3 . (9.33)
With the switch on, the battery operating provides current I which distributes itself
into I1 , I2 . . . I∞ as it flows through an infinite array of resistors each of which is of
strength R ohms.
Physical behavior of this circuit effectively replicates that of Fig. 9.7. The battery
experiences an effective total resistance Reffective and drives current I .
Ampere’s law can now be used to determine the effective total current driven
through the single effective resistance Reffective .
V
Ieffective = . (9.34)
Reffective
An infinitely long repetitive array is not affected if a finite part is extracted.

Therefore, the physics of the circuit in Fig. 9.7 is exactly replicated by that of circuit
in Fig. 9.8. As shown, when current I reaches point A, it splits between two resistors
R and Reffective . From point A to point B, current I1 flows through resistor R, while the
remainder, I −I1 , flows through the resistor Reffective . Both these currents are driven
under the same potential difference. Therefore
I1 R = [I − I1 ] Reffective . (9.35)
Combining this result with (9.33) and (9.34) leads to

Reffective V /R
I1 = I = √ . (9.36)
R + Re f f ective 2+ 3
Next consider an infinite network of resistors shown in Fig. 9.9. The battery oper-
ates at constant voltage V across an on–off switch and a trio of resistors R, R0 , R.
This is the first-trio of this network. The middle resistor of the first-trio, namely resis-
tor R0 , is connected in parallel to an identical second-trio R, R0 , R, and the process
is repeated ad infinitum. Current flow and the effective resistance of are treated next.
In the network shown in Fig. 9.10, current I amps distributes itself into I1 , I2 . . . I∞
as it flows through the infinite array of resistors. In order to work out the current I1 ,
the total effective current Ieffect0 , the total effective resistance Reffect0 , one can follow
the following procedure. Connect the midpoint resistor R0 in parallel to a new resistor
of strength Reffect0 . And, as shown in Fig. 9.10, stop there. The upshot again is that the
resistance of this assembly of four resistors should exactly be equal to the effective
resistance, Reffect0 , of the infinite network. That is,
9.2 Infinite Networks of Resistors 273
Reffect0 = R + R + R0 Reffect0 . (9.37)
The above is a quadratic in Reffect0 with two solutions: one positive and another
un-physical solution that is negative. The physical solution is

R0
Reffect0 = R 1 + 1+2 . (9.38)
R
Some features of the effective resistance, Reffect0√, are worth noting.

First: In the limit R0 → R, Reffect0 → R[1 + 3]. This—according to (9.33)—is
as it should be.
Second: In the limit R0 → ∞, Reffect0 → ∞. Again this makes sense because now
the total network consists of two infinite strings of resistors R that are unconnected—
accept perhaps at ∞ !—and each of the strings has infinite resistance.
Third: The limit R0 → 0, where Reffect0 → 2 R. To understand this result directly
from Fig. 9.9 requires a little visual alacrity. In contrast, Fig. 9.10 is much clearer.
When current I arrives at the point A, it sees two options for crossing over to point
B. First, via zero resistance. Or second, through a finite resistance Reffect0 . Quite
sensibly, all of the current chooses to travel through zero rather than finite resistance.
Hence, Reffect0 is only R + R.
9.2.1 Current Flow
When the switch is on, current I flows out of the positive terminal of the battery
and, after completing its journey, returns to the negative terminal. The total effective
resistance of the circuit is Reffect0 given in (9.38). Therefore, the total effective current
is
V V
I = = . (9.39)
Reffect0 R0
R 1+ 1+2 R
9.2.2 Current I1
When current I reaches the point A in Fig. 9.10, it splits between two resistors R0
and Reffect0 . From point A to point B, current I1 flows through resistor R0 , while the
remainder, I − I1 , flows through resistor Reffect0 . Both these currents flow under the
same potential difference. Therefore
I1 R0 = [I − I1 ] Reffect0 (9.40)
leading to the result

Reffect0 V
I1 = I = . (9.41)
R0 + Re f f ect0
R0 + R 1 + 1 + 2 RR0
Next consider an infinite network of resistors shown in Fig. 9.11. The battery operates
at constant voltage V volts across an on–off switch and a trio of resistors R1 , R2 , R3 .
This is the first-trio. The middle resistor of the first-trio, namely resistor R2 , is
connected in parallel to an identical second-trio R1 , R2 , R3 . And this process is
repeated similarly ad infinitum. The effective resistance and the effective current
flow are treated below.
The requirement that the effective resistance of the infinite circuit be the same as
RNORC leads to the equality
RNORC = R1 + R3 + R2 RNORC . (9.42)
As a quadratic it has two solutions: one of which is positive and the other is negative
and therefore un-physical.

(R1 + R3 ) R1 + R3 2
RNORC = + + R2 (R1 + R3 ) . (9.43)
2 2
Some notable features of RNORC are its results in certain limits.

(1) R1 → R2 → R3 ≡ R—meaning when all resistors are equal.
√
RNORC → R 1 + 3 . (9.44)
This result, according to (9.33), is correct.

(1) : R3 → 0—meaning, when the lower line in the infinite circuit is a perfectly
conducting wire—the result is

2
R1 R1
RNORC = + + R1 R2 . (9.45)
2 2
(2) : When R3 → 0 and R2 → R1 ,
R1 √
RNORC = 1+ 5 . (9.46)
2
9.2.3 Current Flow
When the switch is on, current I amps flows out of the positive terminal of the battery
and, after completing its journey, returns to the negative terminal. The total effective
resistance of the circuit is RNORC —see (9.43). Therefore, the total effective current
is
V
I =
RNORC
V
= . (9.47)
(R1 +R3 )
R1 +R3 2
2
+ 2
+ R2 (R1 + R3 )
[After leaving the positive terminal of the battery, the current I arrives at point A.
There it splits into two parts. Current I1 amps travels across the resistor R2 to point
B. The remainder, (I − I1 ) amps, travels through RNORC . At point B, currents I1
and (I − I1 ) join up so that their sum, namely the current I , returns to the battery.
Because the currents flowing through the resistors, R2 and RNORC , are driven by the
same voltage difference, the following relationship must obtain.
I1 R2 = [I − I1 ] RNORC (9.48)
With the help of (9.47), one can write (9.48) as

RNORC V
I1 = I = . (9.49)
R2 + RNORC R2 + RNORC
Figure 9.13 portrays an infinite network which is a somewhat more involved

network than those considered before. Shown on the left are three resistors in series,
each of strength R1 ohms. These three resistors are referred to as the first-triplet. The
first-triplet and an on–off switch are connected across a large battery that operates at
V volts. Turning the switch on makes it a closed-loop circuit.
The middle resistor R1 of the first-triplet is connected to another three resistors
each of which is of strength R2 ohms. These six resistors make up the first-sextet.
Following the first-sextet is a second-sextet that replicates the first-sextet. The second-
sextet is followed by an identical third-sextet, and this process of adding sextet after
sextet is repeated ad infinitum to form the infinite network. The effective resistance
is—see Fig. 9.14—Reff . The effective total current, Ieff , breaks up into infinite num-
ber of different parts which flow through the given infinite circuit. According to
Ohm’s law, Ieff = RVeff .
9.2.4 Effective Total Resistance
Calculate the effective total resistance of the infinite network drawn in Fig. 9.13. Here
it is helpful to proceed in four successive stages.
To begin with notice the equivalence of Figs. 9.13 and 9.14.
First Stage: Treat the first-sextet as distinct from the rest. Focus on its second-
triplet that is composed of three resistors of strength R2 ohms each. Next, connect
the middle resistor R2 of this second-triplet in parallel to a new resistor of strength
Reff . This process has now formed a septet as shown in Fig. 9.14.
Second Stage: However, as shown in Fig. 9.15, this septet of resistors can again
be reduced further to an equivalent-sextet by representing the last two resistors—
that is, R2 connected in parallel to Reff —by a single resistor R0 . And because of the
parallel nature of this connection, the relevant relationship is
R0 = R2 Reff . (9.50)
Notice this relationship has two unknowns: R0 and Reff . To determine these two
unknowns, one needs also a second relationship. For that purpose, proceed as follows.
Third Stage: The so-named equivalent-sextet achieved in the second stage
described above can be reduced now to an equivalent-triplet. This is done by focussing
on the resistor in the middle of the first-trio of resistors of strength R1 . Because this
resistor is aligned in parallel to three other resistors, R2 , R0 , and R2 , all four can be
replaced by a single equivalent resistance—see Fig. 9.16—Rq . The relevant relation-
ship is
Rq = R1 (R0 + 2 R2 ) . (9.51)
As shown in Fig. 9.16, we are now left with a simple circuit: consisting of the power
source and three resistors—namely, R1 , Rq , and R1 —that are connected in series.
Fourth Stage: Finally replace these three resistors by an equivalent-singlet: a sin-
gle effective resistor, Re f f ect−Cc , that is connected to the power source—see Fig. 9.17.
This means
Re f f ect−Cc = R1 + Rq + R1 . (9.52)
9.2.5 The Result for Re f f ect−C c
There are a total of three equations—namely, (9.50), (9.51), and (9.52)—and three
unknowns: namely, Re f f ect−Cc , R0 , and Rq . After combining (9.50), (9.51), and
(9.52), straightforward algebra is used to calculate Re f f ect−Cc as a function of R1
and R2 . However, even without examining the result in detail, one can predict its
outcome in three particular limits: (1), (2), (3).
(1) Consider the resistance notated Rq . In the limit R2 − > ∞, (9.51) gives
Rq − > R1 . As such (9.52) would lead to
Re f f ect−Cc − > 3 R1 . (9.53)
(2) Next consider R0 . In the opposite limit, namely R2 − > 0, (9.50) leads to
R0 − > 0 . (9.54)
(3) And with R0 and R2 both tending to zero, (9.51) leads to Rq − > 0. As a result
(9.52) yields
Re f f ect−Cc − > 2 R1 . (9.55)
9.2.6 Re f f ect−C c as a Function of R1 and R2
For convenience, introduce the notation

R2
x = . (9.56)
R1
After some straightforward but time-consuming algebra, Re f f ect−Cc can be calculated

by combining (9.50), (9.51), and (9.52). The result is
√
1 + 4x − x 2 + 1 + 10x + 26x 2 + 10x 3 + x 4
Re f f ect−Cc = R1 . (9.57)
(1 + 3x)
It is helpful to have a few simple tests that check the accuracy of (9.57).
(a) : To that end, first try R2 − > ∞. In that limit, according to (9.56), x − > ∞.
Therefore (9.57) transforms as follows.

4x − x 2 + x 2 (1 + 5/x)
Re f f ect−Cc x −>∞
= R1
(3x)

4x − x + x 2 + 5x
2
= R1 = 3R1 . (9.58)
(3x)
This is the result in (1) above.

(b) : Next, look at the case R2 − > 0. That implies x − > 0. In this limit (9.57)
leads to
√
1+ 1
Re f f ect−Cc x −>0
= R1 = 2 R1 . (9.59)
1
(c) : Finally, consider R2 = R1 : that is, x = 1. Setting x = 1 in (9.57) correctly
leads to the result obtained earlier in (9.33).
√
4 + 48 √
Re f f ect−Cc x −>1
= R1 = R1 1 + 3 . (9.60)
4
9.2.7 Table 1 : Electric Circuit Elements
Whenever convenient, some of the following notation may be used. Also, to indicate
time dependence, symbols may be in lower case.

N otation U sed

Capacitance, C Measur ed in f arads : F

Charge, Q Measur ed in coulombs : C

Curr ent, I Measur ed in amper es : A

I nductance, L Measur ed in henries : H

Resistance, R Measur ed in ohms :

V oltage, V Measur ed in volts : V
(9.61)
9.3 Inductors
A coil made of several turns of a very good conducting wire is the physical equivalent
of an inductor (L). If constant current i(0) flows through (L), the coil acts merely as
a very week resistor. According to Ampere’s law, constant current produces a time-
independent magnetic field whose magnitude is proportional to the current. On the
other hand, if the current i(t) is changing in time, it produces a changing magnetic
field. And the resulting change in the magnetic flux induces an electro-motive force
(emf). The (emf) produced is proportional to the rate of change of the current.
Because of the weakness of the internal resistance, current flows that are constant
in time mostly produce un-substantial changes in voltage acrossan inductor.
On the
d i(t)
other hand, when current flow is changing in strength, say at rate dt amperes per
second, an inductor of strength L henrys affects nonzero potential drop v(t) across it.
d i(t)
v(t) = L . (9.62)
dt
9.3 Inductors 279
9.3.1 Two Inductors in Series
Imagine time-dependent current i(t) amps flowing through two inductors of strength
L 1 and L 2 henries connected together in series. The drop in potential across the first
inductor is L 1 di(t)
dt
and that across the second is L 2 di(t)
dt
. And because these drops in
potential occur in series, the total drop in potential is their sum. The same result would
obtain if the given current passed through a single inductor of strength (L 1 + L 2 )
henries.
d i(t) d i(t) d i(t)
L1 + L2 = (L 1 + L 2 ) . (9.63)
dt dt dt
In other words, when placed in series the inductances add much the same way as do
resistances.
9.3.2 Two Inductors in Parallel
Imagine time-dependent current i(t) amps flowing through two inductors of strength
L 1 and L 2 henries connected together in parallel. The drop in potential across the
first inductor is L 1 di(t)
dt
and that across the second is L 2 di(t)
dt
. And because these drops
in potential occur in parallel, they are the same.
d i 1 (t) d i 2 (t)
L1 = L2 ≡ Const . (9.64)
dt dt
With slight manipulation of (9.64), one can write
d i 1 (t) Const
= ;
dt L1
d i 2 (t) Const
= . (9.65)
dt L2
Denote the total current i parll and total inductance L parll and use (9.65) to write
d i parll d i 1 (t) d i 2 (t) Const Const Const
≡ + = + ≡ . (9.66)
dt dt dt L1 L2 L parll
Equation (9.66) leads to the following relationship obeyed by the total inductance,
L parll , of two inductors, L 1 and L 2 , added together in parallel.
1 1 1
= + . (9.67)
L parll L1 L2
Once again, the above is similar to the behavior of resistors.

9.3.3 Infinite Network of Inductors
Consider the infinite network of inductors L 1 , L 2 , L 3 shown in Fig. 9.18. Its effective
inductance may equivalently be represented by the inductance of a single inductor
of strength L eff . As a result, inductance L eff may be equated to the inductance of the
four inductors shown in Fig. 9.19. That is,
L eff = L 1 + L 2 + L 3 L eff . (9.68)
Equivalently
L 2eff = L eff (L 1 + L 2 ) + L 3 (L 1 + L 2 ) . (9.69)
This is a quadratic with a positive and a negative solution. The physically acceptable
one is the positive solution

(L 1 + L 2 ) 4 L3
L eff = 1+ 1+ (9.70)
2 (L 1 + L 2 )
Because in the limit L 3 → 0 points A and B get connected, L eff must → L 1 + L 2 —

exactly as suggested by (9.70).
9.3.4 Capacitors
Two conducting plates, separated by dielectric constitute a parallel plate capacitor

(C). [Note, a good dielectric conducts electricity very poorly but supports electric
fields well.] Equal amounts of opposite charge may be stored on opposing plates
of (C). If these charges are +Q and −Q, and the potential difference between the
opposing plates is V, the capacitance is C.
Q coulombs
C farads = . (9.71)
V volts
As such, a farad is equal to coulombs per volt: or, equivalently, (amperes. second)
per volt. Also, a coulomb is equivalent to farads. volt.
9.3.5 Charging a Capacitor
Assume a large battery that can supply current at fixed voltage V is connected to a
capacitor (C), a perfect resistor (R), and a switch. [Note : A perfect resistor obeys
9.3 Inductors 281
Ohm’s law exactly and contains no stray inductance or capacitance.] These items are
placed in series forming a closed-loop circuit. The charging of (C) requires an inflow
of current into, say, the left plate and an equal outflow from the right plate. Generally
this process is not instantaneous. Rather, the charging occurs exponentially with a
relaxation time that is representative of the nature of the capacitor as well as details
of the resistance in the circuitry. And both the current i as well as the charge q are
functions of time. Note, current and charge are related because one represents the
rate of flow of the other. That is,
dq(t)
i(t) = . (9.72)
dt
Therefore at some specified time t, the charge q(t) that has accumulated on the left
plate is
t t
dq(t)
q(t) = dt = i(t) dt . (9.73)
−∞ dt −∞
Of course, at that time t, the charge on the right plate is −q(t).
9.3.6 Two Capacitors in Parallel
Imagine two capacitors, C1 and C2 , connected in parallel. This implies the left plates
of the two capacitors are connected together as are the right plates. As a result,
both capacitors experience the same voltage difference, say V volts, leading to the
relationships
Q 1 = C1 · V ; Q 2 = C2 · V ; Q 1 + Q 2 = (C1 + C2 ) · V , (9.74)
where Q 1 and Q 2 are the charges held on the positive plates of capacitors C1 and
C2 , respectively. The total charge held is Q total−parallel and the total capacitance is
Ctotal−parallel .
Q total−parallel = Q 1 + Q 2 = (C1 + C2 ) · V ≡ Ctotal−parallel · V . (9.75)
Equation (9.75) leads to the result
Ctotal−parallel = (C1 + C2 ) . (9.76)
This is an interesting result in that it is similar to that of resistors being combined in

series, rather than in parallel.
9.3.7 Two Capacitors in Series
Imagine two capacitors, C1 and C2 , joined together in series. This implies the negative
plate of capacitor C1 is connected to the positive plate of capacitor C2 . As a result, the
charges will be distributed as follows. If the left plate—meaning the positive plate—
of capacitor C1 has charge +Q, its right plate will have charge −Q. This will result
in the left plate of capacitor C2 with charge +Q and its right plate with charge −Q.
And if the potential drop across capacitor C1 is V1 , and that across C2 is V2 , the total
potential drop across the two capacitors in series will be Vtotal−series−2 = V1 + V2 and
the following relationships will hold.
Q Q Q
Vtotal−series−2 = V1 + V2 = + ≡ (9.77)
C1 C2 Ctotal−series−2
leading to the result
1 1 1
= + . (9.78)
Ctotal−series−2 C1 C2
If the two capacitors are equal—that is, C1 = C2 ≡ C—their sum in series is equal
to a half of each: meaning Ctotal−series−2 = C2 .
The above results are similar to those of resistors being added in parallel. An
interesting consequence of this is that when very large number of capacitors are
added together in series, their sum is very small. For instance, the result of adding n
capacitors of strength C in series is Cn , and if n → ∞ the sum is zero.
9.3.8 Infinite Network of Capacitors
Consider the infinite network of capacitors C1 , C2 , C3 shown in Fig. 9.20. Its effec-
tive capacitance may equivalently be represented by the capacitance of a single
capacitor of strength Ceff . As a result, capacitance Ceff may be equated to the capac-
itance of the four capacitors shown in Fig. 9.21.
That is,
1 1 1 1
= + + . (9.79)
Ceff C1 C2 C3 + Ceff
Equivalently
2
Ceff (C1 + C2 ) = − Ceff C3 (C1 + C2 ) + C1 C2 C3 . (9.80)
9.3 Inductors 283
This is a quadratic with a positive and a negative solution. The physically acceptable
solution is the positive one, namely
2
C3 C3 C1 C2 C3
Ceff = − + + . (9.81)
2 2 C1 + C2
9.3.9 Comment
Some of the implications of (9.81) are the following.

(a) In the limit C3 → 0, Ceff → 0.
This result makes eminent sense because now the network is composed only of
infinite numbers of capacitors, C1 and C2 , connected in series. Equivalently, infinite
numbers of resistors connected in parallel would have vanishing resistance.
(b) Essentially similar argument applies to the case when capacitors C1 → 0
C2 .C3
because now the network resembles an infinite array of capacitors C0 ≡ C2 +C3
that are connected in series. Thus, one expects that in the limit C1 → 0, Ceff would
also → 0.
C2 →
(c) Clearly, the same is also true if capacitors 0 because here the network
C1 .C3
resembles an infinite array of capacitors C0 ≡ C1 +C3 that are connected in series.
(d) Finally, if both capacitors C1 and C2 → 0, the network would resemble an
infinite array of C3 capacitors that are connected in series. As a result Ceff → 0.
(e) When all capacitors are equal, meaning C1 = C2 = C3 ≡ C, according to
(9.81) the effective capacitance of the corresponding infinite network in Fig. 9.20:
Circuit (C A P AC I T ) would be Ceff .
C √
Ceff = 3 −1 . (9.82)
2
9.4 R-C Series-Circuit
Figure 9.22 displays a series circuit comprised of an on–off switch, a large battery,
a resistor R, and a capacitor C. At time t the switch is on, the battery supplies time-
dependent current i(t) at constant voltage V, and the charges on the left and the
right plates of the capacitor are q(t) and −q(t). The following differential equation
describes this closed-loop circuit.
q(t) dq(t) q(t)

R i(t) + = V ≡ R + . (9.83)
C dt C
On the right-hand side of (9.83), i(t) has been replaced by dq(t)

dt
. Initially, meaning at
t ≤ 0, the capacitor does not have any charge on either of its plates and both the plates
are at zero potential. However, instantly after the switch is turned on, current starts
flowing, and the process of charging the capacitor gets initiated. As the charge begins
accumulating on the plates—left side positive and the right side negative—potential
difference between the plates begins increasing from its initial value of zero. As a
result, the effective voltage across the resistor begins decreasing. This engenders
time-dependent decrease of the current. The resultant charging of the plates is an
exponential process, and eventually, the current stops flowing, the capacitor gets
fully charged, and the potential difference across the plates of the capacitor equals
the battery voltage V. It is convenient to divide both sides of (9.83) by R and re-write
it as
dq(t) q(t) V
+ = (9.84)
dt τ R−C R
where
τ R−C ≡ R C (9.85)
is the time constant of the R − C series circuit. When R is measured in —ohms—

and C in F—farads—τ R−C is measured in seconds.
Recall that inhomogeneous ordinary differential equations with constant coeffi-
cients and their solutions were discussed in detail in (3.74)–(3.80). And (9.84) is a
simple, one-dimensional such differential equation. Its solution is

t
q(t) = σ0 exp − + CV . (9.86)
τ R−C
The unknown constant σ0 can be determined from one boundary condition.
Problem
Assume the capacitor in the system described above and shown in Fig. 9.22 is in its
uncharged state. Turn the switch on at time t = t0 and work out the time dependence
of the charge and the current.
Solution
Charge on the capacitor at time t0 is zero. Equation (9.86) at t = t0 ,
9.4 R-C Series-Circuit 285

t0
q(t0 ) = 0 = σ0 exp − + CV , (9.87)
τ R−C
gives

t0
σ0 = − C V exp . (9.88)
τ R−C
Accordingly, we have

t − t0
q(t) = C V 1 − exp − (9.89)
τ R−C
and

dq(t) V t − t0
i(t) = = exp − . (9.90)
dt R τ R−C
Equation (9.90) predicts that within a time interval equal to τ R−C the current drops to
about one-third—actually, 36.7879441%— of its original value VR , while, accord-
ing to (9.89), the magnitude of the charge on either plate rises to about two-thirds of
its maximum value, which is C V.
9.4.2 R-L Series Circuit
At time t, the switch is on and the battery supplies time-dependent current i(t) at
constant voltage V to a series connection constituted of a resistor R and an inductor
L. The following differential equation describes this closed-loop circuit.
di(t)
R i(t) + L = V. (9.91)
dt
Divide both sides by L and solve the above. The result is

R V
i(t) = σ0 exp − t + . (9.92)
L R
Consider a closed-loop circuit containing a large battery that can supply current at
constant voltage V. Other items in the circuit are a resistor R, an inductor L, and a
switch. All these items are connected in series. Turn the switch on at time t = t0 and
work out the time dependence of the charge and the current.
9.4.4 Solution
At t = t0 , the current is vanishing. As such the solution (9.92) gives

R V
i(t0 ) = 0 = σ0 exp − t0 + . (9.93)
L R
Or equivalently

V R
σ0 = − exp t0 . (9.94)
R L
Inserting the above value of σ0 into (9.92) yields the result

V (t − t0 )
i(t) = 1 − exp − (9.95)
R τ R−L
where τ R−L is the R − L system time constant.
L
τ R−L = . (9.96)
R
When R is measured in —ohms—and L in H—henrys—τ R−L is measured in
seconds.
interval equal to τ R−L the current drops to 63.212056%
Note that within a time
of its maximum value VR .
9.4.5 L-C Series-Circuit
The following is a differential equation that describes a closed-loop circuit formed

of a large battery that at time t supplies current i(t) at constant voltage V to a series
connection constituted of an inductor (L), a capacitor (C), and a switch that is
on—see Fig. 9.23.
q(t) di(t)
+L = V . (9.97)
C dt
9.4 R-C Series-Circuit 287
Replacing i(t) by dq(t)

dt
and dividing both sides by L lead to a second-order inhomo-
geneous linear ordinary differential equation for the charge q(t) on the left plate of
the capacitor. [The charge on the right plate is −q(t).]
d2 q(t) q(t) V
+ = . (9.98)
dt 2 (τ L−C ) 2 L
In (9.98)
√
(τ L−C ) ≡ LC (9.99)
is the time constant of the L-C series circuit.

√ When L is measured in H and C in F,
τ L−C is measured in seconds. [Note : onehenr y.one f arad= 1 second]
The solution to (9.98)—which is an inhomogeneous linear ordinary differential
equation with constant coefficients—is found in the usual manner [Compare (3.55)
to (3.80)].
q(t) = σ1 sin (ω t) + σ2 cos (ω t) + C V ;

dq(t)
i(t) = = ω [σ1 cos (ω t) − σ2 sin (ω t)] . (9.100)
dt
The charge q(t) and the current i(t) are periodic in time with angular frequency ω
and frequency ν.
1
ω = = 2πν . (9.101)
τ L−C
The unknown constants σ1 and σ2 , in (9.100), can be determined by two boundary

conditions.
Problem
Imagine Fig. 9.23 without the battery. And consider an L-C series circuit that despite
the absence of the battery has an on/off switch. The switch is kept off as long as the
time t is < t0 . The charge ±Q 0 on the plates stays put as is while the switch is off.
When at t = t0 the switch is turned on, it allows the charge to start flowing. Describe
the resultant discharge of the capacitor.
Battery Absent
Solution
In the absence of the battery, the voltage V can be excluded from the L − C series-
circuit differential equation (9.98). The two unknown constants, σ1 and σ2 , in the
result of the remaining equation are determined from the known boundary condition:
q(t0 ) = Q 0 and i(t0 ) = 0.
q(t0 ) = Q 0 = σ1 sin (ω t0 ) + σ2 cos (ω t0 ) ;

i(t0 ) = 0 = ω [σ1 cos(ω t0 ) − σ2 sin(ω t0 )] . (9.102)
Thus
σ1 = Q 0 sin(ω t0 ) ;
σ2 = Q 0 cos (ω t0 ) . (9.103)
Inserting the results given in (9.103) into (9.100) yields
q = Q 0 cos [ω (t − t0 )] ;
i = −ω Q 0 sin [ω (t − t0 )] . (9.104)
Accordingly, at time t, the voltage drop across the capacitor is

q Q0
= cos [ω (t − t0 )] . (9.105)
C C
And that across the inductor L is

di Q0
L = − ω 2 L Q 0 cos [ω (t − t0 )] = − cos [ω (t − t0 )] .
dt C
(9.106)
Because there is no impressed voltage, the sum of these two voltage drops is zero.
9.5 L-R-C Series Circuit
Constant Impressed Voltage
9.5.1 Examples Group IV
Problem
Consider an L-R-C circuit with a battery that operates at constant potential V. The
switch is kept off as long as the time t is < 0. While the switch is off, the capacitor
plates have no charge. When at t = +0 the switch is turned on, it allows charge to
start flowing. Describe the charge q(t) and its rate of accumulation.
9.5 L-R-C Series Circuit 289
Fig. 9.24 R-L-C

series-circuit, with a battery
and on–off switch. At time t,
charges on the capacitor
plates are ±q(t) and current
i(t) flows through the circuit
9.5.2 Solution
The L-R-C series-circuit shown in Fig. 9.24 contains resistance R ohms, inductance
L henries, and capacitance C farads. Shown also is the battery that provides constant
voltage V volts. Also there is an on–off switch. Before the switch is turned on at
t=0, the capacitor is completely uncharged. At t = +0, charge +dq streams out
from the positive terminal of the battery and starts getting deposited on the left-
hand plate of the capacitor. As a result, equal amount of negative charge—that is,
−dq—is induced on the right-hand plate of the capacitor. Because of the physical
requirement that charge neither be created nor destroyed, this process results in
equal amount of positive charge—that is +dq—to move away from the right-hand
plate of the capacitor, continue its travel around the circuit, and in time dt return
to the negative terminal of the battery. This whole process is equivalent to current,
dq
dt
, flowing across the entire circuit, from the positive terminal of the battery back
through to the negative terminal.
Streaming of the charge and changes in potential across the circuit are intercon-
nected. According to Kirchhoff’s second law, over a complete cycle, the total change
in potential is zero. Assuming the battery provides current at V volts, a flowchart of
the voltage changes across the whole circuit, at a given time t ≥ 0, would be as fol-
lows. Voltage change across capacitor C + voltage change across inductor L + volt-
age change across resistor R + voltage change across the battery = 0. That is,
q(t) di(t)
− −L − R i(t) + V = 0 . (9.107)
C dt
Divide (9.107) by L , replace i(t) by dq(t)

dt
, and rearrange slightly.

d2 q(t) R dq(t) 1 V
+ + q(t) = . (9.108)
dt 2 L dt LC L
The above is an inhomogeneous linear ordinary differential equation with constant

coefficients C. Its solution, according to (3.59), is
q(t) = Scomp (t) + I pi (t) (9.109)
where Scomp (t) is the complementary solution and I pi (t) is the particular integral that
are defined as follows.
2
d R d 1
+ + Scomp (t) = 0 . (9.110)
dt 2 L dt LC
2
d R d 1 V
2
+ + I pi (t) = . (9.111)
dt L dt LC L
According to the description given in detail in Chap. 3—see (3.55) and (3.56)—
Scomp (t) is readily calculated. Similarly, I pi (t) may be found by using (3.61) and
(3.67). We get
⎧ ⎫
⎨ R 2 1 ⎬
Rt
Scomp (t) = σ1 exp − exp t −
2L ⎩ 2L LC ⎭
⎡ ⎧ ⎫⎤
⎨ R 2 1 ⎬
Rt
+ σ2 exp − exp ⎣− t − ⎦ . (9.112)
2L ⎩ 2L LC ⎭

V 1
I pi (t) = 1
= CV . (9.113)
L LC

It is interesting to note that in (9.110) and (9.111) , the expression L1C plays the
R
role of angular velocity squared, while L acts as a friction coefficient. [Note : In
(9.112) the constants σ1 and σ2 are arbitrary and, as usual, can be determined by two
boundary conditions.] For convenience, introduce the notation
2
R 1
α = − . (9.114)
2L LC
Using the above notation, re-write (9.112) as
Scomp (t) = σ1 exp ( 1 t) + σ2 exp ( 2 t). (9.115)

where

−R
1 = +α,
2L

−R
2 = −α . (9.116)
2L
No Impressed Voltage
When the battery is absent, the impressed voltage, V, is zero. As a result, I pi (t) is
vanishing and according to (9.109) and (9.115)
q(t) = Scomp (t) = σ1 exp ( 1 t) + σ2 exp ( 2 t). (9.117)
9.5.3 Over-Damped Series Circuit
Capacitor Charge q(t)

The circuit is over-damped when the resistance R is overpowering and, as a result,
α—see (9.114)—is real rather than imaginary. This results in the discriminant being
greater than zero, that is,
2
R 1
α2 = − > 0, (9.118)
2L LC
which leads to another inequality

2
R
> α2 . (9.119)
2L
Because L, R, and C are all positive, (9.119) implies

R
> α (9.120)
2L
and (9.114) indicates that α is positive. Then, according to (9.114), (9.116),

and (9.120), we have the inequalities
1 < 0 ;
2 < 0 ;
1− 2 > 0 . (9.121)
In view of these inequalities, (9.117) predicts that irrespective of whether σ1 and σ2

are positive or negative, as time passes the charge q(t) must exponentially decrease,
heading to zero as t− > ∞.
9.5.4 Current Flow Across Capacitor
Shown in Fig. 9.25 is a closed-loop series circuit that contains a resistor, an inductor,
a capacitor, and an on–off switch. The switch is off while t ≤ 0. At t = 0, the
charge on the left plate of the capacitor is +q(0) and the current i(0) = 0. However,
immediately as t > 0 the switch is turned on and the capacitor begins discharging:
thereby resulting in current i(t) flowing from the positively charged left plate all the
way around the circuit to the negatively charged right plate of the capacitor.
Problem
Work out the time dependence of the capacitor charge q and the electric current i in
the circuit. For numerical calculation use:
I nductance L = 1 H ; Resistance R = 103 ;
−1
Capacitance C = 9 · 104 F ; Capacitor Charge q(0) = 10−3 C
Solution
Clearly I pi (t) = 0 for t ≥ 0 because there is no impressed voltage.
Regarding the charge q(t) on the capacitor, and the current i(t) flowing in the
circuit, at t = 0 we are told i(0) = 0 and according to (9.115)
Scomp (t = 0) = q(0) = σ1 + σ2 = 10−3 C . (9.122)
Fig. 9.25 Similar to

Fig. 9.24 except for the
battery which is missing here
To determine the current i(t), differentiate (9.117).
dq(t)
i(t) = = σ1 1 exp ( 1 t) + σ2 2 exp ( 2 t). (9.123)
dt
Because the current i(0) is vanishing, (9.123) gives
i(0) = σ1 1 + σ2 2 = 0A . (9.124)
Combining (9.122) and (9.124) leads to

10−3 . C
2 2
σ1 = − q(0) = −
1 − 2 1 − 2

10−3 .C
1 1
σ2 = q(0) = (9.125)
1 − 2 1 − 2

According to (9.114) and (9.116), one needs 2RL and L1C for calculating α,
1 , and 2 . Using the numbers provided, one gets

R 1
= 500 ; = 90, 000 ; α = 400 ;
2L LC
= −100 ; 2 = − 900 ;
1
−3 −3
10 10
σ1 = 9 ; σ2 = − . (9.126)
8 8
Therefore, (9.117) and (9.123) give

10−3
q(t) = 9 exp(−100 t) − exp(−900 t) C ;
8

9
i(t) = − exp(−100 t) − exp(−900 t) A . (9.127)
80
Charge Reduction Exponential

The magnitude of the charge q(t) on the plates reduces exponentially as time passes.
The behavior of the current, however, is more interesting (Fig. 9.26).
Current Decrease Exponential

Although it stays negative, it goes through an extremum at t = 0.0025993 s when
the current has reached −0.0759057 A. Thereafter the current strength decreases
exponentially (Fig. 9.27).
Fig. 9.26 An over-damped

L-R-C series circuit: the
capacitor charge q(t) C is
plotted as function of time t
S
Fig. 9.27 Current −i(t) in

an over-damped L-R-C
series circuit as function of
time t. The extremum in − i
is 0.0759 A and it occurs
when t = 0.0026 S
9.5.5 Critically Damped Series Circuit
Capacitor Charge q(t)

If the discriminant—see (9.118)—α2 is vanishing, meaning if

R 2 1
− = 0, (9.128)
2L LC
the two roots of the characteristic equation are equal. That is,

R
1 = 2 ≡ 0 =− . (9.129)
2L
Therefore, according to the well-established procedure—see (3.37)—

(9.117) does not apply. Rather, the charge q(t) must then be expressed in the form
q(t) = (σ3 + σ4 t) exp( 0 t) . (9.130)
As a result, the current i(t) is

dq
i(t) = =[ 0 (σ3 + σ4 t) + σ4 ] exp( 0 t) . (9.131)
dt
9.5.6 Examples Group V
Problem
Assume the capacitor charge is q(0) as the impressed voltage is turned off at t = 0.
Immediately thereafter, the capacitor begins discharging . Work out the time depen-
dence of the electric current and the (magnitude) of the charge on a capacitor plate.
For numerical calculation use :
I nductance L = 1 H ; Resistance R = 2 × 102 ;

−1
Capacitance C = 104 F ; Capacitor Charge q(0) = 10−4 C.
These numbers obey (9.128) which is a requirement for a critically damped circuit.
Solution
At t = 0, the charge is q(0) and the current i(0). Inserting this information into
(9.130) and (9.131), as well as noting the fact that i(0) = 0, we are led to
q(0) = σ3 ,
i(0) = 0 = ω0 σ3 + σ4 (9.132)
which gives
− ω0 q(0) = σ4 . (9.133)
Using the results obtained in (9.132) and (9.133) in (9.130) and (9.131), one gets

R Rt
q(t) = q(0) 1 + t exp − (9.134)
2L 2L
and
2
R R
i(t) = − q(0) t exp − t . (9.135)
2L 2L
Fig. 9.28 L-R-C circuit with

battery. Current i 1 (t) flows
out and on arrival at point A
current i(t) separates and
moves up the inductor L. The
remaining current continues,
eventually joining with i(t)
at point B
Turn the switch on in the L − R − C circuit shown in Fig. 9.28. The large battery
supplies current at constant voltage V volts. At time t, current i 1 (t) is flowing out
of the positive terminal of the battery. At the first circuit point A, current i(t) sepa-
rates from i 1 (t) and moves up the inductor L to the circuit point B. The remaining
current— that is, i 1 (t) − i(t)—continues onward and moves up the capacitor C even-
tually joining at the second circuit point B with the current i(t). Together these two
currents make up current i 1 (t) which, after passing through the resistor R, returns
to the negative terminal of the battery: thereby completing the closed-loop circuit.
Assume at time t, the charge on the lower plate of the capacitor is +q(t) and that on
the higher plate is −q(t).
Problem
Refer to Fig. 9.28 and work out currents i(t), i 1 (t) and charge q(t).
Solution
Figure 9.28 actually describes three separate circuits. Each follows its own differential
equation and makes its own contribution to the process.
For instance, the first circuit on the left that contains the battery, the inductor L,
and the resistor R. Its differential equation is
di(t)
V −L − R i 1 (t) = 0 . (9.136)
dt
Next, there is the L-C circuit that is contained within points A and B. As shown
in Fig. 9.24, both the inductor L and the capacitor C experience identical potential
difference that obtains between points A and B: hence the equality
di(t) q(t)
L = (9.137)
dt C
where
dq(t)
= i 1 (t) − i(t) . (9.138)
dt
In addition to the above three equations, one can also write a fourth equation that
refers to the circuit heading from the battery through the capacitor back to the battery.
That is,
q(t)
V = + R i1 . (9.139)
C
It turns out that (9.139) does not provide any new information beyond what is already
included in the first set of three equations (9.136), (9.137), and (9.138). But that is
not a problem because the first set of equations is sufficient for determining the three
unknowns: i(t), i 1 (t) and q(t). In order to carry out this determination, proceed as
follows.
Differentiate (9.137) and combine the result with (9.138).
d2 i(t) 1 dq(t)
L 2
=
dt C dt
i 1 (t) − i(t)
= . (9.140)
C
Eliminate i 1 (t) from (9.140) and (9.136),

d2 i(t) L di(t) i(t) V
L + + = , (9.141)
dt 2 R C dt C RC
and divide (9.141) by L . This leads to the desired, single, differential equation
whereby i(t) can be determined.

d2 i(t) 1 di(t) 1 V
+ + i(t) = . (9.142)
dt 2 RC dt LC L RC
The above is an inhomogeneous linear ordinary differential equation with constant

coefficients. Its solution, according to (3.59), is
i(t) = Scomp (t) + I pi (t) (9.143)

where Scomp (t) is the complementary solution and I pi (t) is the particular integral that
are defined as follows.
2
d 1 d 1
+ + Scomp (t) = 0 . (9.144)
dt 2 R C dt LC
2
d 1 d 1 V
+ + I pi (t) = . (9.145)
dt 2 R C dt LC L RC
According to the description given in detail in Chap. 3—see (3.55) and (3.56)—
Scomp (t) is readily calculated. Similarly, I pi (t) may be calculated by using (3.61)
and (3.67).
⎧ ⎫
⎨ 1 2 1 ⎬
t
Scomp (t) = σ1 exp − exp t −
2RC ⎩ 2RC LC ⎭
⎧ ⎫
⎨ 2
t 1 1 ⎬
+ σ2 exp − exp −t − . (9.146)
2RC ⎩ 2RC LC ⎭

V 1 V
I pi (t) = 1
= . (9.147)
L RC LC
R
Following (9.137) → (9.147), both the charge q(t) and the current i 1 (t) can now
be calculated. For instance, (9.137) gives
di(t)
q(t) = L C (9.148)
dt
which leads to
⎧ ⎫
⎨ 2 ⎬
−L t 1 1
q(t) = σ1 exp − exp t −
2R 2RC ⎩ 2RC LC ⎭
⎧ ⎫
⎨ 2 ⎬
−L t 1 1
+ σ2 exp − exp −t −
⎡ ⎧ ⎫⎤
⎨ 1 2 1 ⎬
L 2 t
+ − (LC) ⎣σ1 exp − exp t − ⎦
⎡ ⎧ ⎫⎤
⎨ 2 ⎬
L 2 t 1 1
− − (LC) ⎣σ2 exp − exp −t − ⎦.
(9.149)
Fig. 9.29 At time 0, the two

plates of the capacitor hold
charges Q 0 and −Q 0 .
Shown here is
q(t) versus i 1 (t), and i 2 (t)
And knowing q(t) and i(t)—see (9.157), (9.143), (9.146), and (9.147)—(9.138)
provides a simple route to the evaluation of the current i 1 (t).
Problem
At time t = 0, the two plates of the capacitor hold charges q(0) ≡ Q 0 and −Q 0 and
current i 2 (t = 0) is equal to I20 . Work out q(t), i 1 (t), and i 2 (t). The relevant plot is
in Fig. 9.29.
Solution
The potential difference across points A and E is the same as that across points B
and D.
di 2 (t)
R i 1 (t) = L . (9.150)
dt
The potential difference across points B and D is also equal to that across the capac-
itor C.
di 2 (t)
L = q(t)/C . (9.151)
dt
Additionally, the current flow out of the positive plate—and back into the negative
plate—of the capacitor leads to the relationship
dq(t)
= − i 1 (t) − i 2 (t) . (9.152)
dt
Differentiate (9.151) and combine the result with (9.152).
di 2 2 (t) 1 dq(t)
L =
dt 2 C dt
−i 1 (t) − i 2 (t)
= . (9.153)
C
Now eliminate i 1 (t) from (9.153) and (9.150) and divide the result by L to get
d2 i 2 (t) 1 di 2 (t) i 2 (t)

2
+ + = 0 . (9.154)
dt R C dt LC
Much like (9.146), differential equation (9.154) leads to
⎧ ⎫
⎨ 1 2 1 ⎬

t
i 2 (t) = σ1 exp − exp t −
2RC ⎩ 2RC LC ⎭
⎧ ⎫
⎨ 2
t 1 1 ⎬
+ σ2 exp − exp −t − . (9.155)
2RC ⎩ 2RC LC ⎭
According to (9.150) and (9.151)

L di 2 (t)
i 1 (t) =
R dt

1
= q(t) . (9.156)
RC
The relationship L C didt2 (t) ≡ q(t) leads to

⎧ ⎫
⎨ 2 ⎬
−L t 1 1
q(t) = σ1 exp − exp t −
⎧ ⎫
⎨ 2 ⎬
−L t 1 1
+ σ2 exp − exp −t −
⎡ ⎧ ⎫⎤
2 ⎨ 2 ⎬
L t 1 1
+ − (LC) ⎣σ1 exp − exp t − ⎦
⎡ ⎧ ⎫⎤
2 ⎨ 2
L ⎣ t 1 1 ⎬⎦
− − (LC) σ2 exp − exp −t − .
≡ i 1 (t) RC . (9.157)
Equations (9.155) and (9.157) provide the solution to Examples Group 7 except that
there still are two unknowns σ1 and σ2 . The given two boundary conditions, namely
i 2 (t = 0) = I20 and q(t = 0) = Q 0 , determine these unknowns. That is,
I20 = σ1 + σ2 (9.158)
and
2
−L L
Q0 = − (σ1 + σ2 ) + (σ1 − σ2 ) − (LC) . (9.159)
2R 2R
Chapter 10
Numerical Solution
Given a first-order linear differential equation
dy(x)
= F(x, y) ,
dx
and its solution, y(x0 ), at a point x = x0 , Runge–Kutta procedure is used to estimate

y(x0 + ). Runge–Kutta procedure is the least accurate when it uses only one step
for the entire move. And indeed, as noted later, the single-step process does yield
grossly inaccurate results. The two-step process—see (10.12) and (10.13)—improves
the results only slightly. But the four-step
1 effort—see (10.14)–(10.17)—does much
better. It reduces the error to about 50 th of that for the one-step process. Estimates
from a ten-step Runge–Kutta
process are recorded in Table 10.1. These estimates—
being in error only by 100 × 0.002522.17
= 0.0113%—are highly accurate.
Coupled first-order differential equations are treated next.
dx
A(x, y) = =x+y ,
dt
dy
B(x, y) = =x−y . (10.1)
dt
Together these equations are equivalent to a single second-order differential equation.
Tables 10.2, 10.3, 10.4, 10.5, 10.6, and 10.7 display numerical results for the one-step,
two-step, and the five-step processes. Table 10.8 contains numerical results gathered
during a twenty-step Runge–Kutta process. At maximum extension, = 2, the
Runge–Kutta estimate is 26.371190. It differs from the exact result, 26.371404, by
only a tiny amount, 0.000214. The percentage error involved is 0.000811. Table 10.9
records numerical results collected during a twenty-step Runge–Kutta process.
At maximum extension, = 2, the Runge–Kutta estimate for Yn is 11.0170347.
It differs from the exact result, 11.0171222, by 0.0000875. The percentage error
involved is 0.000794. It is similar to the corresponding error, 0.000811%, for X n .

https://doi.org/10.1007/978-3-319-76406-1_10
304 10 Numerical Solution
The accuracy achieved by the twenty-step Runge–Kutta estimate is quite extraor-

dinary. When very high accuracy is desired, the twenty-step Runge–Kutta process
yields results that are worth the effort.
10.1 Single First-Order Differential Equations
10.1.1 Runge–Kutta Steps
Consider a single first-order differential equation.
dy(x)
= F(x, y) . (10.2)
dx
Its solution, y(x), contains one arbitrary constant that can be determined by speci-
fying one boundary condition. Let that boundary condition be the value of y(x) at
x = x0 . That is
y(x0 ) = Y0 . (10.3)
Assume the given differential equation cannot be solved by methods that have been
described so far in this book. The objective of the current exercise is to work out an
approximation procedure whereby one can proceed beyond the starting point in a
step-by-step process. Each step is chosen to be of length . In this manner, n steps
are needed to move from the initial location x0 to the desired final location xn .
xn = x0 + n = x0 + . (10.4)
In other words, knowing y0 = y(x0 ) = Y0 , one attempts to find Yn that is an

approximation to yn = y(xn ) = y(x0 + ). However, before starting, one must
choose numerical values for the distance and the total number of steps, n. As a
result one is led to the single-step length specified by

= . (10.5)
n
The protocol for carrying out this objective is suggested by Runge and Kutta. It
involves successive use of small ‘steps,’ each of which carries the process through a
small extension in position, say from x to (x +). To the above purpose, one proceeds
as follows. The first Runge–Kutta step of length takes us from the starting point
x0 to a neighboring point x1 = x0 + and requires calculating the parameters
R1:1 , R1;2 , R1;3 , R1;4 , K 1;1 . Here Y1 represents an estimate of the exact result y1 =
y(x1 ) = y(x0 + ). Of course, the hope is that the extension, , is short enough that
the estimate Y1 is a good approximation to the unknown exact result y1 .
Choose the value of x0 , Y0 = y(x0 ), n, and . Set = n .
10.1 Single First-Order Differential Equations 305
Runge–Kutta : First Step
= ... ; x0 = ... ; Y0 = ... ;

R1;1
R1;1 = F (x0 , Y0 ) ; R2;1 = F x0 + , Y0 + ;
2 2

R2;1
R3;1 = F x0 + , Y0 + ; R4;1 = F x0 + , Y0 + R3;1 ;
2 2
1
K1 = R1;1 + 2 R2;1 + 2 R3;1 + R4;1 ;
6
Y1 = Y0 + K 1 . (10.6)
It is important to note that Y0 has been chosen to be exactly equal to a given value.
Therefore, it is exact. On the other hand, Y1 is the Runge–Kutta estimated value of
y1 . Therefore, it is approximate. The objective of the current exercise is to make this
approximation as accurate as needed. And an important option for achieving this is
to make short.
Because is short, one needs to extend through several ’s. Usually, the beginning
and the endpoints, say x0 and x0 + , are determined by the physical needs of
the subject matter that is being studied. And, of course, greater the number n, the
greater the effort needed to arrive at the desired estimate. Therefore, the choice for
n is made by the amount of effort one is willing to expand. And because of the
relationship (10.5), that decision fixes the value of .
We could now proceed to the second step. But it is more convenient directly to
go to the nth step.
Runge–Kutta : nth-Step
= ... ; xn−1 = ... ; Yn−1 = ... ;

R1;n
R1;n = F (xn−1 , Yn−1 ) ; R2;n = F xn−1 + , Yn−1 +
2 2

R2;n
R3;n = F xn−1 + , Yn−1 + ; R4;n = F xn−1 + , Yn−1 + R3;n
2 2
1
Kn = R1;n + 2 R2;n + 2 R3;n + R4;n ;
6
Yn = Yn−1 + K n . (10.7)
10.1.2 Runge–Kutta Solution
The Runge–Kutta procedure is best demonstrated by working out an actual problem.

To that purpose, consider the following first-order differential equation.
dy(x)
= xy . (10.8)
dx
Assume we do not know how to solve it. Because it is first order, its solution must
contain one arbitrary constant that can be determined by a single boundary condition.
Let such boundary condition be its solution at x = x0 = 0 and assume that this
solution is equal to Y0 = 3. That is
y(x = x0 = 0) = Y0 = 3 . (10.9)
The Objective
Knowing x0 and y(x0 ), we wish to find the solution y(x) at a point x = x0 + .
Need for Exact Solution
As mentioned earlier, in principle we do not know the exact solution of (10.8).
For educational purposes, however, it is important to know how well the Runge–
Kutta procedure actually is faring. To that end, it is best to have available, hidden
somewhere in the background, the exact solution. Then, as we proceed through
the various Runge–Kutta steps, the accuracy of the numerical approximation can
properly be evaluated.
The exact boundary-value solution of (10.8) is

x2 x2
y(x) = a0 exp = 3 exp . (10.10)
2 2
Solution with Longest Possible

In order to make our work the shortest possible, we need to use the longest-possible
: meaning = = 2. Then, working through the first step described in (10.6) we
have
= 2 ; x0 = 0 ; Y0 = 3 ;

R1;1
R1;1 = x0 Y0 = 0 ; R2;1 = x0 + Y0 + = 6;
2 2

R2;1
R3;1 = x0 + Y0 + = 12 ;
2 2

R4;1 = (x0 + ) Y0 + R3;1 = 60 ;
1
K1 = R1;1 + 2 R2;1 + 2 R3;1 + R4;1 = 16 ;
6
Y1 = Y0 + K 1 = 3 + 16 = 19 . (10.11)
The Y1 calculated above is the estimated value of y(x = ). Its exact value is
2
y(x = 2) = 3 exp( 22 ) = 22.167168. Thus, for the longest-possible choice for , the
Runge–Kutta estimate is quite

inaccurate. It equals 19. Therefore, it is in error by
3.17 points which is about 100 × 22.17
3.17
= 14.3%.
Solution When Is Half-the-Longest Possible
Next, let us deal with the case where is equal to half-the-longest possible.
Given is equal to one-half the maximum-possible length—meaning
= 2 = 1—one needs to walk through two successive Runge–Kutta steps. For the
first step, one has [Note: Compare (10.6)]
= 1 , x0 = 0 , Y0 = 3 ;
R1;1 = 0 ; R2;1 = 1.5 ;
R3;1 = 1.875 ; R4;1 = 4.875
1
K1 = R1;1 + 2 R2;1 + 2 R3;1 + R4;1 = 1.9375 ;
6
Y1 = Y0 + K 1 = 3 + 1.9375 = 4.9375 . (10.12)
And for the second step
= 1 ; x1 = ; Y1 = 4.9375 ;
R1;2 = e x1 Y1 = 4.9375 ; R2;2 = 11.1094 ;
R3;2 = 15.7383 ; R4;2 = 41.3516 ;
1
K2 = R1;2 + 2 R2;2 + 2 R3;2 + R4;2 = 16.6641 ;
6
Y2 = Y1 + K 2 = 4.9375 + 16.6641 = 21.6016 . (10.13)
Y2 = 21.6016 , calculated above, is the estimated value of y(2). Its exact value is
3 exp(2) = 22.167168. Thus, for half-the-longest-possible choice for , the Runge–
Kutta estimate is already much improved over that for the longest-possible choice.
The result is now only 0.57 points in error, which is less than a fifth of the previous
error.
Solution with One-Fourth the Longest Possible
Next, let us consider an even shorter value of equal to a quarter of the longest-
possible choice: meaning = 0.5. This requires trudging through four Runge–Kutta
steps, each of length . The first step yields
= 0.5 ; x0 = 0 ; Y0 = 3 ;
R1;1 = 0 ; R2;1 = 0.375 ;
R3;1 = 0.398438 ; R4;1 = 0.849608 ;
1
K1 = R1;1 + 2 R2;1 + 2 R3;1 + R4;1 = 0.399414 ;
6
Y1 = Y0 + K 1 = 3 + 0.399414 = 3.399414. (10.14)
The second step gives
= 0.5 ; x1 = ; Y1 = 3.399414 ;
R1;2 = 0.849853 ; R2;2 = 1.43413 ;
R3;2 = 1.54368 ; R4;2 = 2.47154 ;
1
K2 = R1;2 + 2 R2;2 + 2 R3;2 + R4;2 = 1.54617 ;
6
Y2 = Y1 + K 2 = 3.399414 + 1.54617 = 4.945584. (10.15)
The third step gives
= 0.5 ; x2 = 2 ; Y2 = 4.945584 ;
R1;3 = 2.47279 ; R2;3 = 3.86373 ;
R3;3 = 4.2984 ; R4;3 = 6.93299 ;
1
K3 = R1;3 + 2 R2;3 + 2 R3;3 + R4;3 = 4.28834 ;
6
Y3 = Y2 + K 3 = 4.945584 + 4.28834 = 9.233924 (10.16)
The fourth- and the final step leads to
= 0.5 ; x3 = 3 ; Y3 = 9.233924 ;
R1;4 = 6.92544 ; R2;4 = 11.1096 ;
R3;4 = 12.9401 ; R4;4 = 12.174 ;
1
K4 = R1;2 + 2 R2;2 + 2 R3;2 + R4;2 = 12.8665 ;
6
Y4 = Y3 + K 4 = 9.233924 + 12.8665 = 22.100424 (10.17)
Equation (10.17) indicates that the use of one-quarter the longest-possible leads
to Runge–Kutta approximation result, 22.100424.
This result
is in error by a mere
22.167−22.100 = 0.067 points. Being only 100 × 22.1670.067
= 0.302%, this is greatly

reduced error. Indeed, it is almost one-fiftieth of 100 × 22.167−19
22.167
= 14.29%—
which was the error when the longest-possible was used.
Solution with One-Tenth the Longest-Possible
To get a good feel as to whether the improvement in the estimate is worth the addi-

tional effort, let us try even a shorter length : say, = 10 = 0.2. In order to save
space, these results are tabulated in Table 10.1.
Table 10.1 indicates that when the travel to the maximum extension, x10 = 2 = ,

is carried out by using = 10 —meaning, when ten steps are used to get to the
maximum extension— the Runge–Kutta estimate is in error ≈0.0025. The percentage
error now—being about 0.0113%—is absolutely minuscule. Indeed it is much less
than one-thousandth the error, 14.3%, found by using the longest possible value of
.
Table 10.1 Ten-step xn Exact Runge–Kutta Error

numerical approximation
results 0. 3 3 0
0.2 3.060604 3.060604 0.0
0.4 3.249861 3.249861 0.0
0.6 3.591652 3.591651 0.000001
0.8 4.131383 4.131378 0.000005
1.0 4.946164 4.946149 0.000015
1.2 6.163260 6.163255 0.00005
1.4 7.993369 7.993240 0.00013
1.6 10.789919 10.789563 0.00036
1.8 15.159271 15.158317 0.00096
2.0 22.167168 22.164669 0.0025
Summing-Up: Results for One-Tenth the Longest-Possible

‘For an initial value first-order differential equation, Runge–Kutta provides excellent
approximation to the exact result when ten Runge–Kutta steps are used. Indeed, using
less than half this effort—meaning when one-fourth the longest-possible is used–
the resultant Runge–Kutta estimate is in error by only 0.302% which may often be
acceptable accuracy.
10.2 Coupled Differential Equations First Order
Second-order differential equations can be constructed from coupled first-order dif-

ferential equations. For instance, consider the following differential equations:
dx
A(x, y) = =x+y ,
dt
dy
B(x, y) = =x−y . (10.18)
dt
Differentiate with respect to t the upper (10.18) and write the result as follows.
d2x dx dy dx dx dx
2
= + = +x−y = +x+x−
dt dt dt dt dt dt
= 2x . (10.19)
Similarly, differentiate the lower (10.18) and write
d2 y dx dy dy dy dy
2
= − = x+y− = +y+y−
dt dt dt dt dt dt
=2y. (10.20)
The second-order differential equations (10.19) and (10.20) can straightforwardly

be solved by the usual, trial solution as exponentials, technique. Meaning set x(t) =
const. exp(α t) and y(t) = const. exp(β t). In this manner, the solutions are
√ √
x(t) = a1 exp ( 2 t) + b1 exp (− 2 t) ,
√ √
y(t) = a2 exp ( 2 t) + b2 exp (− 2 t) . (10.21)
While we expect the solution to a first-order differential equation to contain only a

single arbitrary constant, unfortunately the given solutions—see (10.21)— to two
first-order differential equations seem to involve a total of four such constants. These
constants are a1 , b1 , a2 , b2 . Fortunately, by using the procedure described in (3.28)–
(3.37), one can properly reduce these constants to the needed total of only two. The
result is
√ √
x(t) = a1 exp ( 2 t) + b1 exp (− 2 t) ,
√ √ √ √
y(t) = 2 − 1 a1 exp ( 2 t) − 2 + 1 b1 exp (− 2 t) . (10.22)
And the two constants a1 and b1 can be determined by the two boundary conditions—
namely
x(t = t0 = 0) = 1
= a1 + b1 = X 0 ;
y(t = t0 = 0) = 2
√ √
= 2 − 1 a1 − 2 + 1 b1 = Y0 . (10.23)
Equation (10.23) lead to

√ √
3+ 2 2−3
a1 = √ ; b1 = √ . (10.24)
2 2 2 2
Inserting the a1 and b1 —given in (10.24)—into (10.22) provides the desired exact
solution of (10.18).
√ √
3+ 2 √ 2−3 √
x(t) = √ exp ( 2 t) + √ exp (− 2 t) ;
2 2 2 2
√ √
3+ 2 √
y(t) = 2−1 √ exp ( 2 t)
2 2
√ √
2−3 √
− 2+1 √ exp (− 2 t) . (10.25)
2 2
10.2 Coupled Differential Equations First Order 311
Equation (10.25) duly satisfy the required two boundary conditions (10.23). And they
provide ready access to the exact result for comparison with Runge–Kutta estimate.
One hastens to add that the Runge–Kutta numerical approximation—to be dis-
cussed in the following—for a coupled pair of first-order differential equations can
be employed without knowing the exact results in (10.25). The latter are provided
merely for educational purposes. In other words, exact results help evaluate the accu-
racy of the Runge–Kutta estimate.
10.2.1 Runge–Kutta Steps
Use the information = 0.1, t0 = 0, x(t0 ) = X 0 = 1, y(t0 ) = Y0 = 2 and note

the following:
X n is the estimated value of xn = x(tn ) = x(t0 + n ), Yn is the estimated value
of yn = y(tn ) = y(t0 + n ), and t0 has been set equal to zero.
The two coupled first-order differential equations that need to be solved—subject
to the two boundary conditions (10.23)—are A(x,y) and B(x,y) as given in (10.18).
The nth Runge–Kutta step refers to tn = t0 + n with t0 = 0 and xn and yn are
exact results.
10.2.2 Numerical Solution
One-Step Numerical Approximation

Solution with the Longest-Possible
Let us start with the worst possible choice for : meaning equal to the farthest
separation envisioned. That is = = 2. Remember, X 0 = 1 and Y0 = 2 and the
Runge–Kutta objective is to move to X 0 + = 1 + 2 = 3 and Y0 + = 2 + 2 = 4.
The results are provided in Tables 10.2 and 10.3. In the single-step process, the
Runge–Kutta estimate for final X n is in error by 4.7. The exact value is 26.37. In
4.7
percentage terms, for the one-step process, the error in final X n is 100 26.37 =
0.35
17.8%. The corresponding error in final value of Yn is 100 11.02 = 3.18%.
Table 10.2 = 2: one-step n Exact Runge–Kutta Error

approximation for X n
0 1. 1. 0.
2 26.371404 21.66667 4.7
Table 10.3 = 2: one-step n Exact Runge–Kutta Error

approximation for Yn
0 2. 2. 0.
2 11.017122 10.666667 0.35
Relationships and Results for Runge–Kutta First Step

The objective at the first step is to calculate X 1 and Y1 which are estimates for the
corresponding exact values x1 = x(t1 ) = x(t0 + ) and y1 = y(t1 ) = y(t0 + ).
Note t0 = 0. The Runge–Kutta equation and the relevant numerical values for the
first step are given below.
R1;1 = (X 0 + Y0 ) = 0.1 × (1 + 2) = 0.3 ;

S1;1 = (X 0 − Y0 ) = 0.1 × (1 − 2) = − 0.1 ;

R1;1 S1;1 0.3 0.1
R2;1 = X 0 + + Y0 + = 0.1 1 + +2− = 0.31
2 2 2 2

R1;1 S1;1 0.3 0.1
S2;1 = X 0 + − Y0 − = 0.1 1 + −2+ = −0.08
2 2 2 2

R2;1 S2;1 0.31 0.08
R3;1 = X 0 + + Y0 + = 0.1 1 + +2− = 0.3115
2 2 2 2

R2;1 S2;1 0.31 0.08
S3;1 = X 0 + − Y0 − = 0.1 1 + −2+ = −0.0805
2 2 2 2

R4;1 = X 0 + R3;1 + Y0 + S3;1 = 0.1 [1 + 0.3115 + 2 − 0.0805] = 0.3231

S4;1 = X 0 + R3;1 − Y0 − S3;1 = 0.1 [1 + 0.3115 − 2 + 0.0805] = −0.0608
1
K1 = R1;1 + 2 R2;1 + 2 R3;1 + R4;1 = 0.311017;
6
1
L1 = S1;1 + 2 S2;1 + 2 S3;1 + S4;1 = − 0.0803;
6
X 1 = X 0 + K 1 = 1.311017 ; Y1 = Y0 + L 1 = 1.919700 ;
x1 = 1.311018 ; y1 = 1.919700 . (10.26)
Detailed information about the second step is provided next.

Relationships and Results for Runge–Kutta Second Step
R1;2 = (X 1 + Y1 ) = 0.1 × (1.311018 + 1.9197) = 0.323072 ;

S1;2 = (X 1 − Y1 ) = 0.1 × (1.311018 − 1.9197) = − 0.060868 ;

R1;2 S1;2
R2;2 = X 1 + + Y1 + = 0.336182
2 2

R1;2 S1;2
S2;2 = X 1 + − Y1 − = − 0.0416713
2 2

R2;2 S2;2
R3;2 = X 1 + + Y1 + = 0.337797
2 2

R2;2 S2;2
S3;2 = X 1 + − Y1 − = − 0.0419756
2 2

R4;2 = X 1 + R3;2 + Y1 + S3;2 = 0.352654

S4;2 = X 1 + R3;2 − Y1 − S3;2 = − 0.022891
1
K2 = R1;2 + 2 R2;2 + 2 R3;2 + R4;2 = 0.337281;
6
1
L2 = S1;2 + 2 S2;2 + 2 S3;2 + S4;2 = − 0.041842;
6
X 2 = X 1 + K 2 = 1.648298; Y2 = Y1 + L 2 = 1.877858 ;
x2 = 1.648299 ; y1 = 1.877857 . (10.27)
The second step refers to t2 = t0 + 2 = 0.2. X 2 is the estimated value of x2 . And

Y2 is the estimated value of y2 . Note, t0 has been set equal to zero.
If the desired total number of Runge–Kutta steps is n, one must proceed step-
by step starting at the first step. The structure of the nth Runge–Kutta step is the
following.
Expected Structure of Runge–Kutta nth Step
R1;n = (X n−1 + Yn−1 ) ;

S1;n = (X n−1 − Yn−1 ) ;

R1;n S1;n
R2;n = X n−1 + + Yn−1 +
2 2

R1;n S1;n
S2;n = X n−1 + − Yn−1 −
2 2

R2;n S2;n
R3;n = X n−1 + + Yn−1 +
2 2

R2;n S2;n
S3;n = X n−1 + − Yn−1 −
2 2

R4;n = X n−1 + R3;n + Yn−1 + S3;n

S4;n = X n−1 + R3;n − Yn−1 − S3;n
1
Kn = R1;n + 2 R2;n + 2 R3;n + R4;n
6
1
Ln = S1;n + 2 S2;n + 2 S3;n + S4;n
6
X n = X n−1 + K n ; Yn = Yn−1 + L n
xn = −... ; yn = ... (10.28)
Two-Step Numerical Approximation

Solution with Half-the-Longest-Possible
Consider next the case = 1 = 2 . Keep X 0 = 1, Y0 = 2, X 0 + = X 0 + 2 =
1 + 2 = 3 and Y0 + = Y0 + 2 = 2 + 2 = 4. The results for X n and Yn are
recorded in Tables 10.4 and 10.5. Here the Runge–Kutta estimate for final X n is
Table 10.4 = 1: two-step n Exact Runge–Kutta Error

0 1. 1. 0.
1 6.28308 6.166667 0.1164
2 26.371404 25.583335 0.7881
Table 10.5 = 1: two-step n Exact Runge–Kutta Error

0 2. 2. 0.
1 3. 3.0 0.
2 11.0171222 10.722223 0.2949
Table 10.6 = 0.4: five-step n Exact Runge–Kutta Error

0 2. 2. 0.
0.4 2.429344 2.428267 0.001
0.8 4.657032 4.653827 0.003
1.2 8.415135 8.407175 0.008
1.6 14.938662 14.920358 0.020
2 26.371404 26.331396 0.040
Table 10.7 = 0.4: five-step n Exact Runge–Kutta Error

0 2. 2. 0.
0.4 1.906948 1.9072 0.0002
0.8 2.440566 2.440032 0.0005
1.2 3.776214 3.773594 0.0026
1.6 6.352821 6.345752 0.0071
2 11.017122 11.000915 0.016207
in error by 0.7881. The exact value is 26.37. In percentage terms, for the two-step
Runge–Kutta process, the error in final value of X n is 100 0.7881
26.37
= 2.9884%. That
is about one-sixth the error, 17.8, for the single-step process.
Regarding the final Yn , the relevant Tables 10.3 and 10.5. The single-step estimate
for Yn is in error by 0.3505 and the two-step processes are in error by 0.2949. In
other words, while the error involved in the single-step process is 3.18%, the two-step
process is in error by 2.68%. The difference between the two sets of errors is not
significant. It is only 0.5%. Perhaps things will improve if we go to a five-step process.
Five-Step Numerical Approximation
Solution with One-Fifth the Longest-Possible
Error Comparison : One-Step Versus Five-Step Process
Examine Table 10.6. At full extension the Runge–Kutta estimate for X n is
26.331396. Compared with the exact result, 26.371404, the estimate is in error
by 0.04008. This error is about ( 120
1
)th the error, 4.705, found by using the one-step
Runge–Kutta process. In percentage terms, the error here is approximately 0.152%.
Table 10.8 = 0.1: n Exact Runge–Kutta Error

twenty-step approximation
0 1. 1. 0.
for X n
0.1 1.311018 1.311017 0.000001
0.2 1.648299 1.648298 0.000001
0.3 2.0186022 2.018599 0.000003
0.4 2.429344 2.429340 0.000004
0.5 2.888754 2.888748 0.000006
0.6 3.406036 3.406027 0.000011
0.7 3.991552 3.991540 0.000012
0.8 4.657032 4.657016 0.000016
0.9 5.415807 5.415787 0.000020
1.0 6.283080 6.283054 0.000026
1.1 7.276224 7.276192 0.000033
1.2 8.415135 8.415094 0.000041
1.3 9.722630 9.722578 0.000052
1.4 11.224901 11.224837 0.000064
1.5 12.952045 12.951966 0.000079
1.6 14.938662 14.938565. 0.000097
1.7 17.22455 17.224431 0.000112
1.8 19.855504 19.855359 0.000128
1.9 22.884230 22.884054 0.000147
2.0 26.371404 26.371190 0.000214
Consider Yn next. The improvement in percentage error is quite dramatic. Accord-

ing to Table 10.7, at full extension the exact result, 11.017122, and the Runge–Kutta
estimate, 11.000915, differ by 0.016207.
1 As such the error in the Runge–Kutta esti-
1
mate after five steps is 0.3505 = 21.63 th that for the one-step process. In percentage
( 0.016207 )
terms, the error in Yn , after five Runge–Kutta steps, is 0.016207 11.00092
× 100 = 0.147.
As such, after five Runge–Kutta steps, the percentage error in Yn is about the same
as that in X n .
Considering the limited amount of effort involved in working through only five
Runge–Kutta steps, the resultant error of only about 0.15%—in estimates for both
X n and Yn —is reasonable. However, if one needs to pursue much greater accuracy,
one might try a twenty-step Runge–Kutta process.
Twenty-Step Numerical Approximation
Solution with One-Twentieth the Longest-Possible
Refer to Table 10.8 for numerical results for X n gathered during a twenty-step Runge–
Kutta process. At maximum extension, = 2, the Runge–Kutta estimate for X n
is 26.371190. It differs from the exact result, 26.371404, by only a tiny amount,
0.000214. The percentage error involved is 0.000811.
Table 10.9 records numerical results Yn collected during a twenty-step Runge–
Kutta process. At maximum extension, = 2, the Runge–Kutta estimate for Yn is
Table 10.9 = 0.1: n Exact Runge–Kutta Error

twenty-step approximation
0 2. 2. 0.000000
for Yn
0.1 1.919700 1.919700 0.000000
0.2 1.877857 1.877858 0.000001
0.3 1.873635 1.873635 0.000001
0.4 1.906948 1.906947 0.000001
0.5 1.978463 1.978462 0.000001
0.6 2.089613 2.089612 0.000001
0.7 2.242626 2.242623 0.000003
0.8 2.440566 2.440561 0.000005
0.9 2.687398 2.687392 0.000006
1.0 2.988068 2.988059 0.000009
1.1 3.348599 3.3485877 0.000011
1.2 3.776214 3.776199 0.000015
1.3 4.279479 4.279460 0.000019
1.4 4.868477 4.868452 0.000025
1.5 5.555006 5.554975 0.000031
1.6 6.352821 6.352782 0.000039
1.7 7.277904 7.277856 0.000048
1.8 8.348788 8.348729 0.000059
1.9 9.586926 9.586854 0.000072
2.0 11.0171222 11.0170347 0.0000875
11.0170347. It differs from the exact result, 11.0171222, by 0.0000875. The percent-
age error involved is 0.000794. It is similar to the corresponding error, 0.000811%,
for X n . The accuracy achieved by the twenty-step Runge–Kutta estimate is quite
extraordinary. When high accuracy is essential, the twenty-step process is well worth
the additional effort.
Chapter 11
Frobenius Solution
11.1 Normalized Form
For A(x) = 0, the normalized form of differential equation
A(x) y (x) + B(x) y (x) + C(x) y(x) = 0 (11.1)
is
y (x) + M (x) y (x) + N (x) y(x) = 0 , (11.2)
where
B(x) C(x)
M (x) = ; N (x) = . (11.3)
A(x) A(x)
11.1.1 An Analytic Function
A function F(x) is analytic at x = x0 if the Taylor series,

∞ dn F

dxn
F(x) = (x − x0 ) n
, (11.4)
n=0
n!
converges to F(x) for all x at, and in the neighborhood of, x0 . Note that geometric
functions, polynomials, and exponentials of analytic functions are analytic as are
rational functions except at the points where their denominator is vanishing.

https://doi.org/10.1007/978-3-319-76406-1_11
318 11 Frobenius Solution
11.1.2 Ordinary Point
If both coefficients, M (x) and N (x), are analytic at x = x0 the point x0 is called an
‘Ordinary Point’ of the differential equation (11.2).
11.1.3 Regular Singular Point
On the other hand, if either or both of these coefficients is/are not analytic at x = x0 ,
but both (x − x0 ) M (x) and (x − x0 )2 N (x) are analytic at x = x0 , then x0 is a ‘Regular
Singular Point’ of the differential equation (11.2).
11.1.4 Irregular Singular Point
But if one and/or other of the products (x − x0 ) M (x) and (x − x0 )2 N (x) is/are not
analytic at x = x0 , the x0 is an ‘Irregular Singular Point’ of the given differential
equation.
11.1.5 Solution Around Ordinary Point
All solutions—compare, Whittaker, E. T. and W atson30. —of a given differential

equation, say (11.2), at or immediately around its ordinary point, say x = x0 , are
analytic and may be expressed as Taylor series such as (11.4). Note, the radius of
convergence of these solutions is equal to at least the separation between x0 and the
nearest singular point of the given differential equation.
11.1.6 Equations of Type (a)
In the following, we shall treat differential equations of type (a)—see (11.5)—around

their ordinary point x = 0 with initial conditions u(x = 0) = σ1 and du dx x=0
= σ2 .
2
(a) : D + α x D + (β x2 + γ) u(x) = 0 , (11.5)
11.1 Normalized Form 319
11.1.7 Solution
In (11.5), the parameters α, β, γ are constants that are real. We classify these equa-
tions as being of type (a). Clearly x = 0 is an ordinary point of the differential
equation (11.5). [Note: The symbols M (x) and N (x) are defined
in (11.2)]. Because
both the relevant coefficients M (x) = α x and N (x) = β x2 + γ are analytic at
x = 0. Frobenius solution of a differential equation around its ordinary point, x = 0,
is found in simple form. Simple form refers to the case where the unknown constant
ν0 is set qual to 0, and the Frobenius solution is expressed as a Maclaurin27. series.
∞

u(x) = an xn . (11.6)
n=0
In order to use (11.6), the differentials that would be needed in (11.5) are the
following:
∞
∞

D2 u(x) = n(n − 1) an xn−2 = n(n − 1) an xn−2 ;
n=0 n=2
∞
∞

Du(x) = n an xn−1 = n an xn−1 . (11.7)
n=0 n=1
Accordingly, the Frobenius expansion of differential equation (11.5) is

∞
∞
∞

n(n − 1) an xn−2 + α x n an xn−1 + (β x2 + γ) an xn = 0 . (11.8)
n=2 n=1 n=0
Unfortunately (11.8), in its current form, is not convenient for working out the details
of the unknown parameters an . To relieve this inconvenience, one would want to
achieve two things. First: All infinite sums should have the same range. Second: All
infinite sums should post the same power of x, say xn .
To begin this process consider, the first term on the left-hand side in (11.8) and
manipulate it as follows:
∞
∞

n(n − 1) an xn−2 = 2 a2 + 6 a3 x + n(n − 1) an xn−2
n=2 n=4
∞

= 2 a2 + 6 a3 x + (n + 2)(n + 1) an+2 xn . (11.9)
n=2
∞
∞ In order to render then term n=4 n(n − 1) an x
n−2
[Note: into

n=2 (n + 2)(n + 1) an+2 x , set n − 2 = n so that n = n + 2. Then write
∞
n=4 n(n − 1) an xn−2 as ∞ n
n +2=4 (n + 2)(n + 1) an +2 x and finally change the

notation n → n to get the desired result.]
Next look at the second term on the left-hand side in (11.8) and manipulate it so
that the infinite sum, like (11.9), is summed from n = 2 and has xn in it. Fortunately
that is easy to achieve.
∞
∞

αx n an xn−1 = α a1 x + α n an xn . (11.10)
n=1 n=2
Finally consider the last term on the left-hand side in (11.8). One gets
∞
∞

an x(n+2) = an −2 xn . (11.11)
n=0 n =2
And another change of variable—this time from n → n—does the trick.

∞
∞
∞

(β x + γ)
2
an x = γ
n
an x + β
n
an x(n+2)
n=0 n=0 n=0
∞ ∞
= γ a0 + γ a1 x + γ an xn + β an−2 xn . (11.12)
n=2 n=2

[Note: To see how β ∞ n=0 an x
(n+2)
becomes β ∞ n
n=2 an−2 x , set n = n + 2 and

change variable from n → n .]
The original differential equation (11.8) is written as: [(11.9)] + [(11.10)] +
[(11.12)] = 0. That is
∞
∞

[2 a2 + 6 a3 x + (n + 2)(n + 1) an+2 x ] + [α a1 x + α
n
n an xn ]
n=2 n=2
∞
∞

+[γ a0 + γ a1 x + γ an xn + β an−2 xn ] = 0 . (11.13)
n=2 n=2
Or more conveniently as
(γ a0 + 2 a2 ) + x(α a1 + γ a1 + 6 a3 )
∞

+ xn {(n + 2)(n + 1) an+2 + (α n + γ) an + β an−2 } = 0 . (11.14)
n=2
Equation (11.14) should be true for all values of x. And that can happen only if the
coefficient multiplying any given power of x is zero: meaning, only if the following
relationships hold:
(γ a0 + 2 a2 ) = 0 , (α a1 + γ a1 + 6 a3 ) = 0 , (11.15)
and for n ≥ 2,
(n + 2)(n + 1) an+2 + an (α n + γ) + β an−2 = 0 . (11.16)
Equations (11.15) and (11.16) lead to

γ α+γ
a2 = − a0 ; a3 = − a1 , (11.17)
2 6
and for all n ≥ 2
an (α n + γ) + β an−2
an+2 = − . (11.18)
(n + 2)(n + 1)
Before proceeding further, it is important to recognize that the differential equation

being solved here is second order and it should have two arbitrary constants for which
a0 and a1 would clearly serve well. Constants a2 and a3 are already known in terms
of these two. All other constants, namely a4 and higher, have to be extracted from
(11.18) and represented in terms of a0 and a1 . That requires beginning with n = 2 in
(11.18).

(2 α + γ) a2 + β a0 (2 α + γ) γ2 − β
a2+2 = − = a0 . (11.19)
12 12
Next use n = 3 in (11.18).

(3 α + γ) a3 + β a1 (3 α + γ) α+γ −β
a2+3 = − = 6
a1 . (11.20)
20 20
With n = 4 (11.18) gives

⎡ γ ⎤
(2 α+γ)( γ2 )−β
(4 α + γ) a4 + β a2 (4 α + γ) 12
−β 2
a6 = − = −⎣ ⎦ a0 .
30 30
(11.21)
⎡ ⎤
(3α+γ) α+γ
6 −β α+γ
⎢ (5α + γ) 20 −β 6 ⎥
(5α + γ) a5 + β a3 ⎢ ⎥
a7 = − = −⎢
⎢
⎥ a1
⎥
42 ⎣ 42 ⎦
(11.22)
Finally, the initial conditions must be satisfied. That is
du
u(x = 0) = σ1 = a0 ; = σ2 = a1 . (11.23)
dx x=0
Inserting (11.17) and (11.19)–(11.23) into (11.6) leads to the result—see (11.24)
given below—for the simple form of the Frobenius series solution of (11.5).

γ (2 α + γ) γ2 − β 4
u(x) = 1 − x + 2
x σ1
2 12
⎧⎡ ⎤ ⎫
⎨ (4 α + γ) (2 α+γ)12( 2 )−β − β γ2
γ
⎬
− ⎣ ⎦ x6 σ1 + O(x8 ) σ1
⎩ 30 ⎭

α+γ 3 (3 α + γ) α+γ −β 5
+ x− x + 6
x σ2
6 20
⎧⎡ ⎤ ⎫
⎨ (5 α + γ) (3 α+γ)20( α+γ
6 )−β
− β α+γ ⎬
6
− ⎣ ⎦ x7 σ2 + O(x9 ) σ2 .
⎩ 42 ⎭
(11.24)
One good thing about having arbitrary constants, α, β, γ, in (11.5) is that different
choices of these constants lead to different differential equations. And its solution in
the form of (11.24) makes the first several terms of their Frobenius series solution
instantly available.
To see how the above process works, let us solve the following five differential
equations which are similar in form to (11.5) which is of type (a).
I − (A) : [D2 + 2 x D + (3 x2 + 4)]u(x) = 0 ,

I − (B) : [D2 + x D + (2 x2 + 1)]u(x) = 0 ,
I − (C) : [D2 + 2 x D + (x2 − 1)]u(x) = 0 .
I − (D) : [D2 + 3 x D + (2 x2 + 1)]u(x) = 0 ,
I − (E) : [D2 − x D + (−x2 − 1)]u(x) = 0 . (11.25)
All we have to do to solve the (11.25) is to use the values of α, β, and γ that are
given. First few terms of the relevant Frobenius power series solution of equations I-
(A)–I-(E) are the following.
For differential equation I-(A), use: α = 2, β = 3, γ = 4. Therefore, its solution:
13 4 7
I − (A) : u(x) = σ1 1 − 2x2 + x − x6 + O(x8 )
12 30
7 19
+ σ2 x − x 3 + x5 − x7 + O(x9 ) . (11.26)
20 420
For I-(B) use: α = 1, β = 2, γ = 1, therefore
1 2 1 29
I − (B) : u(x) = σ1 1 − x − x4 + x6 + O(x8 )
2 24 720
1 3 1 13
+ σ2 x − x − x5 + x7 + O(x9 ) . (11.27)
3 30 630
For I-(C) use: α = 2, β = 1, γ = −1, therefore
1 2 5 23
I − (C) : u(x) = σ1 1 + x − x4 + x6 + O(x8 )
2 24 720
1 3 1 29
+ σ2 x − x − x5 + x7 + O(x9 ) .
6 120 5040
(11.28)
For I-(D) use: α = 3, β = 2, γ = 1, therefore
1 2 1 4 1
I − (D) : u(x) = σ1 1 − x + x − x6 + O(x8 )
2 8 48
2 3 7 2
+ σ2 x − x + x5 + − x7 + O(x9 ) . (11.29)
3 30 35
For I-(E) use: α = −1, β = −1, γ = −1, therefore
1 2 5 37
I − (E) : u(x) = σ1 1 + x + x4 + x6 + O(x8 )
2 24 720
1 3 7 31
+ σ2 x + x + x5 + x7 + O(x9 ) .
3 60 1260
(11.30)
I .1 : [D2 + x D + (−x2 + 1)]u(x) = 0 .

I .2 : [D2 + x D + (−3 x2 + 2)]u(x) = 0 ,
I .3 : [D2 + x D + (3 x2 − 2)]u(x) = 0 ,
I .4 : [D2 + 4 x D + (−x2 − 4)]u(x) = 0 .
I .5 : [D2 + x D + (−x2 + 1)]u(x) = 0 .
I .6 : [D2 + x D + (x2 + 1)]u(x) = 0 ,
I .7 : [D2 − x D − (x2 + 1)]u(x) = 0 ,
I .8 : [D2 + 2 x D + 2 (x2 + 1)]u(x) = 0 .
I .9 : [D2 − 2 x D − 2 (x2 + 1)]u(x) = 0 .
I .10 : [D2 + 3 x D + 3 (x2 − 1)]u(x) = 0 . (11.31)
11.1.10 Equations of Type (b)
Having solved Frobenius equations of Type (a) around their ordinary point, somewhat
more complicated such equations—to be referred to as equations of Type (b)—are
tackled below.
2
(b) : D + (α x2 + β x + γ) D + (μ x2 + ν x + ρ) u(x) = 0 . (11.32)
11.1.11 Solution

Again the initial conditions are: u(x = 0) = σ1 and du dx x=0
= σ2 . All the six con-
stants α, β, γ, μ, ν, ρ are real.Clearly x = 0 is an ordinary point
of (11.32) because
both the coefficients M (x) = α x2 + β x + γ and N (x) = μ x2 + ν x + ρ are an-
alytic at x = 0. Frobenius solution of a differential equation around its ordinary point
x = 0 is best found in simple form. Simple form refers to the case where the unknown
constant ν0 is set qual to 0, and the Frobenius solution is expressed as a Maclaurin
series.
∞

u(x) = an xn . (11.33)
n=0
Using (11.33), the relevant differentials for (11.32) (b) are:

∞
∞

D2 u(x) = n(n − 1) an xn−2 = n(n − 1) an xn−2 ;
n=0 n=2
∞
∞

Du(x) = n an xn−1 = n an xn−1 . (11.34)
n=0 n=1
Accordingly, the Frobenius version of differential equation (b) is
2
D + (α x2 + β x + γ) D + (μ x2 + ν x + ρ) u(x)
∞ ∞

= n(n − 1) an x n−2
+ (α x + β x + γ)
2
n an xn−1
n=2 n=1
∞

+ (μ x2 + ν x + ρ) an xn = 0 . (11.35)
n=0
Unfortunately (11.35), in its current form, is not convenient for working out the
details of the unknown parameters an . To relieve this inconvenience, one would want
to achieve two things. First: All infinite sums should have the same range. Second:
All infinite sums should post the same power of x, say xn .
To begin this process, consider the first term on the right-hand side in (11.35) and
manipulate it as follows:
∞
∞

n(n − 1) an xn−2 = 2 a2 + 6 a3 x + n(n − 1) an xn−2
n=2 n=4
∞

= 2 a2 + 6 a3 x + (n + 2)(n + 1) an+2 xn .
n=2
∞ ∞
n=4 n(n − 1) an x n=2 (n + 2)
n−2
[Note: In order to render the term into

(n
∞ + 1)a n+2 x n
, set n − 2 =
n so that n = n + 2. Then write
as ∞

n=4 n(n − 1) an x n +2=4 (n + 2)(n + 1) an +2 xn and finally change the
n−2

∞
notation n → n to achieve the given result n=2 (n + 2)(n + 1) an+2 xn .]
Therefore, the first term on the right-hand side in (11.35) is
∞
∞

n(n − 1) an xn−2 = 2 a2 + 6 a3 x + (n + 2)(n + 1) an+2 xn . (11.36)
n=2 n=2
Next manipulate the second term on the right-hand side in (11.35) and arrange it so
that the sum incudes xn and ranges from n = 2 to ∞.
∞
∞

(α x2 + β x + γ) n an xn−1 = α (n − 1) an−1 xn + β a1 x
n=1 n=2
∞
∞

+β n an xn + γ n an xn−1 . (11.37)
n=2 n=1
The following transformations were used in (11.37). The first term on the left-hand
side of (11.37) is
∞
∞

α x2 n an xn−1 = α n an xn+1 (11.38)
n=1 n=1
Set n + 1 = n . As a result, the first term on the left-hand side of (11.37) becomes
∞
∞

α x2 n an xn−1 = α n an xn+1
n=1 n=1
∞
∞

=α (n − 1) an −1 xn = α (n − 1) an−1 xn . (11.39)
n =2 n=2
∞
The second term is β x n=1 n an xn−1 . Write it as
∞
∞
∞

βx n an xn−1 = β n an xn = β a1 x + β n an xn . (11.40)
n=1 n=1 n=2
∞
The third term is γ n=1 n an xn−1 . Set n − 1 = n . As a result,
∞
∞
∞

γ n an xn−1 = γ (n + 1) an +1 xn = γ (n + 1) an+1 xn
n=1 n =0 n=0
∞

= γ a1 + 2 γ a2 x + γ (n + 1) an+1 xn . (11.41)
n=2
Consequently, the second term on the right-hand side in (11.35) is

∞

(α x2 + β x + γ) n an xn−1
n=1
∞
∞

=α (n − 1) an−1 xn + β a1 x + β n an xn
n=2 n=2
∞

+ γ a1 + 2 γ a2 x + γ (n + 1) an+1 xn . (11.42)
n=2
Finally, consider the last term on the right-hand side in (11.35).

∞

(μ x2 + ν x + ρ) an xn
n=0
∞
∞ ∞

= μ an x n+2
+ν an x n+1
+ρ an xn . (11.43)
n=0 n=0 n=0
∞
By settingn = n + 2 and changing variable from n → n the sum n=0 an xn+2
becomes ∞ n
n =2 an −2 x . Therefore,
∞
∞
∞

μ an xn+2 = μ an −2 xn = μ an−2 xn . (11.44)
n=0 n =2 n=2
∞ ∞
Similarly by setting n = n + 1, the sum n=0 an xn+1 becomes n =1 an −1 xn .
Therefore,
∞
∞
∞

ν an xn+1 = ν an −1 xn = ν a0 x + ν an−1 xn , (11.45)
n=0 n =1 n=2
and
∞
∞

ρ an xn = ρ a0 + ρ a1 x + ρ an xn . (11.46)
n=0 n=2
By adding (11.44), (11.45), and (11.46), the last term on the right-hand side in (11.35),
that is the current (11.43), becomes
∞
∞

(μ x2 + ν x + ρ) an xn = μ an−2 xn + ν a0 x
n=0 n=2
∞
∞
+ν an−1 xn + ρ a0 + ρ a1 x + ρ an xn . (11.47)
n=2 n=2
In view of the foregoing differential equation (11.35)—which is the Frobenius version

of equation(b)—is written as: [(11.36)] + [(11.42)] + [(11.47)] = 0. That is
2
D + (α x2 + β x + γ) D + (μ x2 + ν x + ρ) u(x)
∞ ∞

= n(n − 1) an xn−2 + (α x2 + β x + γ) n an xn−1
n=2 n=1
∞

+(μ x2 + ν x + ρ) an xn = 0
n=0
∞
∞

= 2 a2 + 6 a3 x + (n + 2)(n + 1) an+2 xn + α (n − 1) an−1 xn
n=2 n=2
∞
∞

+β a1 x + β n an xn + γ a1 + 2 γ a2 x + γ (n + 1) an+1 xn
n=2 n=2
∞
∞
∞

+μ an−2 xn + ν an−1 xn + ν a0 x + ρ a0 + ρ a1 x + ρ an xn .
n=2 n=2 n=2
Or more conveniently as (11.48) given below.

2
D + (α x2 + β x + γ) D + (μ x2 + ν x + ρ) u(x) = 0
= (ρ a0 + 2 a2 + γ a1 ) + x(β a1 + ρ a1 + 6 a3 + 2γ a2 + ν a0 )
∞

+ xn {(n + 2)(n + 1) an+2 }
n=2
∞

+ xn {γ(n + 1)an+1 + (β n + ρ) an + [α(n − 1) + ν] an−1 + μ an−2 } . (11.48)
n=2
Equation (11.48) should hold for all values of x. And that can happen only if the
coefficient multiplying any given power of x is zero: meaning, only if the following
three equations, numbered (11.49), (11.50), (11.51), hold true. [Note: (11.49) refers
to x0 , (11.50) to x1 , and (11.51) to xn for all n ≥ 2.]
(ρ a0 + 2 a2 + γ a1 ) = 0 , (11.49)
(β a1 + ρ a1 + 6 a3 + 2γ a2 + ν a0 ) = 0 , (11.50)
(n + 2)(n + 1) an+2 = − γ(n + 1)an+1 − (β n + ρ) an

− [α (n − 1) + ν] an−1 − μ an−2 , f or all n ≥ 2 . (11.51)
Before proceeding further, it is important to recognize that the differential equation

being solved here is second order and should have two independent constants. Un-
fortunately, (11.49) and (11.50) have four such constants: a0 , a1 , a2 , a3 . If a2 and a3
are eliminated, only two—that is, a0 and a1 —will survive. And then the rest of an
can be written in terms of these two. To that end, proceed as follows.
According to (11.49) and (11.50), we have
γ a1 + ρ a0
a2 = − (11.52)
2
and
(β + ρ) a1 + ν a0 + 2 γ a2 (β + ρ) a1 + ν a0 − γ(ρ a0 + γ a1 )
− =− = a3 .
6 6
Therefore, (11.50) gives
(β + ρ − γ 2 ) a1 + (ν − γ ρ)a0
a3 = − . (11.53)
6
Now that the variables a2 and a3 are available in terms of a0 and a1 , the higher-order
terms—namely a4 , a5 , . . . , etc.—may also be represented as functions of a2 and
a3 . To that purpose, setting n = 2 in (11.51) will lead to a4 , and n = 3 will yield a5 ,
and so on.
μ a0 + (α + ν) a1 + (2 β + ρ) a2 + 3 γ a3
a4 = − . (11.54)
12
Next use n = 3 in (11.51).
μ a1 + (2 α + ν) a2 + (3 β + ρ) a3 + 4 γ a4
a5 = − . (11.55)
20
With n = 4 (11.51) gives
μ a2 + (3 α + ν) a3 + (4 β + ρ) a4 + 5 γ a5
a6 = − . (11.56)
30
For n = 5, (11.51) leads to
μ a3 + (4 α + ν) a4 + (5 β + ρ) a5 + 6 γ a6
a7 = − . (11.57)
42
Finally, the initial conditions must be satisfied. That is
du
u(x = 0) = σ1 = a0 ; = σ2 = a1 . (11.58)
dx x=0
Combining (11.52) to (11.58) with (11.33) provide first few terms for the simple
form of the Frobenius series solution for (11.32).
Use the information given in (11.52)–(11.58) above and solve the following five
differential equations similar in form to equation (b) (11.32). For simplicity, only the
indices α - ρ are given below. [See (11.59)]
II − (F) : [α = 1 ; β = 1 ; γ = 1 ; μ = 1 ; ν = 1 ; ρ = 1 ] .
II − (G) : [α = 0 ; β = 1 ; γ = 2 ; μ = 3 ; ν = 4 ; ρ = 5 ] .
II − (H ) : [α = 1 ; β = 2 ; γ = 3 ; μ = 4 ; ν = 5 ; ρ = 6 ] .
II − (I ) : [α = 2 ; β = 1 ; γ = 2 ; μ = 3 ; ν = 2 ; ρ = 2 ] .
II − (J ) : [α = 1 ; β = 3 ; γ = 3 ; μ = 1 ; ν = 1 ; ρ = 2 ] . (11.59)
Addition of (11.52)–(11.58) readily yields solutions of (11.59)-(F)–(J)— in terms

of a0 ≡ σ1 and a1 ≡ σ2 . Results are given below.
For II-(F) , α = 1 ; β = 1 ; γ = 1 ; μ = 1 ; ν = 1 ; ρ = 1. Therefore,
x2 x4 x5 x6 x7
II − (F) : u(x) = σ1 1 − + + − −
2 24 15 720 70
x2 x3 7 21 43
+ σ2 x − − + x5 + x6 − x7 + O(x8 ) .
2 6 120 720 5040
For II-(G), α = 0 ; β = 1 ; γ = 2 ; μ = 3 ; ν = 4 ; ρ = 5. Therefore,
5 2 17 4 11 5
II − (G) : u(x) = σ1 1 − x + x3 +x − x
2 24 60
5 43 1 5
+σ1 − x6 − x 7 + σ2 x − x 2 − x3 + x4
144 504 3 12
5 6 7
x x x
+σ2 + − + O(x8 ) .
60 72 42
For II-(H), α = 1 ; β = 2 ; γ = 3 ; μ = 4 ; ν = 5 ; ρ = 6. Therefore,
13 3
13 4 23 5
II − (H ) : u(x) = σ1 1 − 3x2 + x +
x − x
6
24 40
103 6 212 3 2 1 3
+σ1 − x − x 7 + σ2 x − x + x
720 5040 2 6
5 4 3 44 335
+σ2 x − x5 − x6 − x7 + O(x8 ) .
8 20 720 5040
For II-(I), α = 2 ; β = 1 ; γ = 2 ; μ = 3 ; ν = 2 ; ρ = 2. Therefore,
1 3 x4 x5
II − (I ) : u(x) = σ1 1 − x2 + x − +
3 12 4
6
x 30 1 3 x4
+σ1 − − x 7 + σ2 x − x 2 + x −
18 1008 6 12
17 x6 23
+σ2 x5 + − x7 + O(x8 ) .
120 40 1008
For II-(J), α = 1 ; β = 3 ; γ = 3 ; μ = 1 ; ν = 1 ; ρ = 2. Therefore,
5 3 x4 17 5
II − (J ) : u(x) = σ1 1 − x2 + x − − x
6 24 60
x6 323 3 2 2 3 x4
+σ1 + x 7 + σ2 x − x + x +
12 5040 2 3 3
6
47 x 516
+σ2 − x5 + + x7 + O(x8 ).
120 720 5040
Use the information given in (11.52)–(11.58) and solve the following five differential
equations similar in form to equation (b) (11.32). For simplicity, only the indices α
to γ are given in (11.60) below.
II .1 : [α = 1 ; β = 2 ; γ = 1 ; μ = 3 ; ν = 3 ; ρ = 1 ] .
II .2 : [α = 0 ; β = 1 ; γ = 0 ; μ = 2 ; ν = 2 ; ρ = 0 ] .
II .3 : [α = 1 ; β = 0 ; γ = 1 ; μ = 2 ; ν = 1 ; ρ = 1 ] .
II .4 : [α = 0 ; β = 2 ; γ = 2 ; μ = 0 ; ν = 0 ; ρ = 1 ] .
II .5 : [α = 2 ; β = 1 ; γ = 1 ; μ = 5 ; ν = 1 ; ρ = 4 ] . (11.60)
11.2 Frobenious Solution Around Regular Singular Point
According to an established idea6. and a recent proo f 30. , if the differential equation
y (x) + M (x) y (x) + N (x) y(x) = 0 (11.61)
has a ‘Regular Singular Point,’ say at x = x0 , then it has at least one solution that is
expressible as a modified Taylor series of the form
∞

y(x) = an (x − x0 )(n+ν0 ) . (11.62)
n=0
The unknown constant ν0 is determined by substituting such series as solution to the

given differential equation.
11.2.1 Equations of Type (c)
Equations similar to (11.63) that is given below will henceforth be referred to as

the equations of Type (c). Here, as in Type (a) and Type (b) equations, all the given
constants—namely β, γ, μ, ν, ρ—are real.
βx+γ μ x2 + ν x + ρ
(c) : y (x) + y (x) + y(x) = 0 . (11.63)
x x2
11.2.2 Solution
Clearly while M (x),
βx+γ
M (x) = , (11.64)
x
and N (x),
μ x2 + ν x + ρ
N (x) = , (11.65)
x2
11.2 Frobenious Solution Around Regular Singular Point 333
are not analytic at x = 0, the products, x M (x) and x2 N (x),
x M (x) = (β x + γ) ,

x N (x) = μ x2 + ν x + ρ ,
2
(11.66)
are analytic at x = 0. As such x = 0 is a regular singular point. In accord with the

Frobenius statement, the differential equation (11.63) is assured to have at least one
solution that can be represented in the form of Maclaurin series (11.67) given below.
∞

y(x) = an x(n+ν0 ) . (11.67)
n=0
In order to solve the second-order differential equation (c) (11.63), the following two
differentials of equation (11.67) are needed.
∞

y (x) = an (n + ν0 ) x(n+ν0 −1) ,
n=0
∞

y (x) = an (n + ν0 )(n + ν0 − 1) x(n+ν0 −2) . (11.68)
n=0
Combining (11.63), (11.67), and (11.68) gives

∞

an (n + ν0 )(n + ν0 − 1) x(n+ν0 −2)
n=0
∞

βx+γ
+ an (n + ν0 ) x(n+ν0 −1)
x n=0
∞

μx + ν x + ρ
2
+ an x(n+ν0 ) = 0 . (11.69)
x2 n=0
Equations (11.63) and (11.69) can be put together in a more useful format as (11.70).
This fact is owed to the Frobenius statement that a solution of (11.63) exists as a
Maclaurin series of the form (11.67).
∞

an [(n + ν0 )(n + ν0 + γ − 1) + ρ ] x(n+ν0 −2)
n=0
∞

+ an [β (n + ν0 ) + ν] x(n+ν0 −1)
n=0
∞

+ [μ] an x(n+ν0 ) = 0 . (11.70)
n=0
The first term of (11.70) can be written as

∞

an {(n + ν0 )(n + ν0 + γ − 1) + ρ } x(n+ν0 −2)
n=0
!
= a0 x(ν0 −2) ν02 + ν0 (γ − 1) + ρ
!
+ a1 x(ν0 −1) ν02 + ν0 (γ + 1) + γ + ρ
∞
+ an {(n + ν0 )(n + ν0 + γ − 1) + ρ } x(n+ν0 −2) . (11.71)
n=2
The second term of (11.70) can be transformed as follows. Set n = n − 1 and then
write the second term of (11.70) as
∞

an [β (n + ν0 ) + ν] x(n+ν0 −1)
n=0
∞

= an −1 [β (n − 1 + ν0 ) + ν] x(n +ν0 −2) . (11.72)
n =1
And finally change the variable n back to the variable n.

∞

an [β (n + ν0 ) + ν] x(n+ν0 −1)
n=0
∞
= an−1 [β (n − 1 + ν0 ) + ν] x(n+ν0 −2)
n=1
∞

= a0 (β ν0 + ν) x(ν0 −1) + an−1 [β (n − 1 + ν0 ) + ν] x(n+ν0 −2) . (11.73)
n=2
The third term of (11.70) can also be suitably transformed. To that end, set n = n + 2
and proceed as follows.
∞
∞

[μ] an x(n+ν0 ) = [μ] an −2 x(n +ν0 −2)
n=0 n =2
∞
= [μ] an−2 x(n+ν0 −2) . (11.74)
n=2
By adding these terms, the original differential equation (11.63) is represented more
neatly as (11.75) given below.
11.2 Frobenious Solution Around Regular Singular Point 335
!
a0 x(ν0 −2) ν02 + ν0 (γ − 1) + ρ
!
+ x(ν0 −1) a1 ν02 + ν0 (γ + 1) + γ + ρ + a0 (β ν0 + ν)
∞
+ an {(n + ν0 )(n + ν0 + γ − 1) + ρ } x(n+ν0 −2)
n=2
∞

+ an−1 [β (n − 1 + ν0 ) + ν] x(n+ν0 −2)
n=2
∞

+ an−2 [μ] x(n+ν0 −2) = 0 . (11.75)
n=2
If (11.75) is to hold for all x, coefficient of every power of x must be vanishing.
11.3 Indicial Equation
Indicial Equation is the Term with the Lowest Power of x in (11.75)
In order to properly satisfy the requirement that (11.75) holds true—namely that
coefficients of every power of x in (11.75) is vanishing—we begin with the first term
of (11.75) which has the lowest power of x—that is, the term a0 x(ν0 −2) —and set its
coefficient equal to zero.
!
ν02 + ν0 (γ − 1) + ρ = 0 . (11.76)
Equation (11.76), being the term that multiplies the lowest power of x in (11.75),
is labeled the ‘Indicial Equation.’
11.4 Indicial Equation Roots
11.4.1 ν1 and ν2
The indicial equation always refers to a specified differential equation. In this case,
the indicial equation (11.76) is particularized to the γ and ρ that appear in differential
equation (11.63), or equivalently (11.75). And being a quadratic, it has two roots, ν1
and ν2 .
Indicial Equation Roots Signify Possible Solutions of the Relevant Differen-

tial Equation
"
2
γ−1 γ−1
ν0 = ν1 = − + −ρ ,
2 2
"
2
γ−1 γ−1
ν0 = ν 2 = − − −ρ . (11.77)
2 2
The two roots of the indicial equation recorded in (11.77), namely ν0 = ν1 and ν0 =
ν2 , signify the two possible solutions of the original differential equation (11.63).
Category (1): Roots Differing by Non-integer

If the two roots, ν1 and ν2 , of the indicial equation are unequal and their dif-
ference is not an integer, then the relevant second-order differential equation
always posseses two linearly independent solutions, y1 (x) and y2 (x).
In this regard, consider the differential equation (11.78) given below.
y (x) + y (x) + y(x) = 0 . (11.78)
x − x0 (x − x0 )2
It has a regular singular point at x = x0 . And should its indicial equation have two
roots that are unequal and differ by an integer. Meaning, if ν1 and ν2 are the two
roots that are unequal and
"
2
γ−1
| ν1 − ν 2 | = 2 | −ρ | , (11.79)
2
is not an integer, then it will always posses two linearly independent solutions, y1 (x)
and y2 (x), in the form of modified Taylor series:
∞

y1 (x) = an (x − x0 )(n+ν1 ) . (11.80)
n=0
∞

y2 (x) = an (x − x0 )(n+ν2 ) . (11.81)
n=0
11.4 Indicial Equation Roots 337
Category (2): Roots Equal

First Solution Readily Found
If the two roots—ν1 and ν2 —of the indicial equation are equal—meaning if
2
ν1 = ν2 ≡ ν0 or, equivalently, ρ = γ−1 2
—and the relevant differential equation—
meaning (11.78)—has a regular singular point at x = x0 , then it will always posses a
relatively easily accessible first solution, y1 (x), in the form of modified Taylor series.
∞

y1 (x) = an (x − x0 )(n+ν0 ) . (11.82)
n=0
Second Solution May Require More Effort

The second solution, y2 (x), in contrast, may require more effort. This issue will be
discussed later in greater detail.
Category (3): Roots Differing by an Integer
If the two roots, ν1 and ν2 , of the indicial equation are unequal—say, with ν1 > ν2 —
and differ by an integer—meaning ν1 − ν2 is integral—and the relevant differential
equation (11.78) has a regular singular point at x = x0 , then there always is a solution,
say y1 (x), that is of the form of a modified Taylor series.
∞

y1 (x) = an (x − x0 )(n+ν1 ) . (11.83)
n=0
The second solution, say y2 (x) that uses ν2 instead of ν1 , may sometime diverge
because of the presence of denominators that go to zero. Often, however, the arbitrary
constant multiplying the whole solution may be changed to obtain non-divergent
second solution. The details of such procedure will become clear when relevant
solutions are discussed in what follows.
Term with the Second Lowest Power of x in (11.75)
Having set the term with the lowest power of x in (11.75) equal to zero, we consider
next the second term—which carries the second lowest power of x, i.e., x(ν0 −1) —and
set it equal to zero.

a1 ν02 + ν0 (γ + 1) + γ + ρ + a0 (β ν0 + ν) = 0 . (11.84)
Equation (11.84) gives

# $
β ν0 + ν β ν0 + ν
a1 = −a0 ! = −a0 . (11.85)
ν0 + ν0 (γ − 1) + ρ + 2 ν0 + γ
2 2 ν0 + γ
Remaining Three Terms of (11.75)

Finally, consider the remaining
three terms of (11.75). They can be set together as a
single sum of the form ∞ n=2 [...] x
(n+ν0 −2)
.
∞

[ {an (n + ν0 )(n + ν0 + γ − 1) + ρ }
n=2
+ an−1 {β (n − 1 + ν0 ) + ν } + an−2 {μ}] x(n+ν0 −2) = 0 .
In order for this sum to add up identically to zero, terms multiplying each and every
power of x must individually tend to zero. Therefore, the following equality
an−1 [β (n − 1 + ν0 ) + ν] + an−2 [μ]

an = − (11.86)
(n + ν0 )(n + ν0 + γ − 1) + ρ
must obtain for all n ≥ 2.

Points to note: Equation (11.85) gives a1 in terms of a0 . Similarly, setting n =
2, 3, 4, 5, 6, 7 in (11.86) readily relates a7 , a6 , a5 , a4 , a3 , a2 to a0 . Finally, unless
otherwise stated, the parameter μ is being set equal to zero. This reduces the total
number of parameters actively being used here to four.
Indicial Equation Roots Unequal and Differing by Non-integer

Equations (11.88) describe a set of ten differential equations of the form
βx+γ νx+ρ
y (x) + y (x) + y(x) = 0 . (11.87)
x x2
These equations are of type (c) where the parameter μ has been dropped. Note that
these equations—(1) to (10)—have all a regular singular point at x = 0. Use (11.85).
Set n = 2, 3, 4, 5, 6, 7 in (11.86). Thereby solve (11.88) (1) to (10). The relevant
four parameters—namely β, γ, ν, ρ—are the following
III − (1) : [β = 1 ; γ = 1 ; ν = 4/3 ; ρ = −1/9 ] .

III − (2) : [β = 0 ; γ = −1/2 ; ν = −1/2 ; ρ = −5/2 ] .
III − (3) : [β = 3 ; γ = −1 ; ν = 1 ; ρ = −5 ] .
III − (4) : [β = 0 ; γ = −1/2 ; ν = 1/2 ; ρ = −5/2 ] .
III − (5) : [β = −1 ; γ = 5/3 ; ν = −2 ; ρ = −1/3 ] .
III − (6) : [β = 1 ; γ = 5/3 ; ν = 2 ; ρ = −1/3 ] .
III − (7) : [β = 1 ; γ = −3/2 ; ν = 1/2 ; ρ = 1 ] .
III − (8) : [β = −1 ; γ = −3/2 ; ν = −1/2 ; ρ = 1 ] .

III − (9) : [β = −1 ; γ = 1 ; ν = −4/3 ; ρ = −1/9 ] .
III − (10) : [β = 3 ; γ = −3 ; ν = 1 ; ρ = −4 ] . (11.88)
All the ten differential equations in (11.88) belong to category (1). Meaning that the
two roots of their indicial equation are different, and the difference between the two
roots is not an integer.
11.4.3 Solution
The solution to examples group III is worked out as follows.

Begin by treating equation III-(1) and keep track of the relevant parameters: that
is β = 1; γ = 1; ν = 4/3; ρ = −1/9. Record the resultant differential equation.
# 1$
x+1 4
x−
y (x) + y (x) + 3 9
y(x) = 0 .
x x2
Consider the indicial equation as described in (11.76).

!
ν02 + ν0 (γ − 1) + ρ = ν0 2 + ν0 (1 − 1) − (1/9) = 0 .
Note that the two roots of the indicial equation given above are ν0 ≡ ν1 = (1/3) and
ν0 ≡ ν2 = (−1/3). They are different and their difference is not an integer. Work
first with ν0 ≡ ν1 = (1/3) and use (11.85) that relates a1 to a0 .
# $
β ν0 + ν 1
+ 4
a1 = −a0 = − a0 3 3
= − a0 . (11.89)
2 ν0 + γ 2
3
+1
To calculate a2 and higher-order an , for convenience, rewrite (11.86) in the form of

(11.90) given below.
β (n − 1 + ν0 ) + ν
f or n ≥ 2 , an = − an−1 . (11.90)
(n + ν0 )(n + ν0 + γ − 1) + ρ
For n = 2, (11.90) gives
(4/3) + (4/3) a1 a0
a2 = − a1 = − = . (11.91)
(7/3)(7/3) − (1/9) 2 2!
(7/3) + (4/3) a2 a0
a3 = − a2 = − = − . (11.92)
(10/3)(10/3) − (1/9) 3 3!
For n = 4, (11.90) gives
(10/3) + (4/3) a3 a0
a4 = − a3 = − = . (11.93)
(13/3)(13/3) − (1/9) 4 4!
For n = 5, (11.90) gives
(13/3) + (4/3) a4 a0
a5 = − a4 = − = − . (11.94)
(16/3)(16/3) − (1/9) 5 5!
(16/3) + (4/3) a5 a0
a6 = − a5 = − = . (11.95)
(19/3)(19/3) − (1/9) 6 6!
For n = 7, (11.90) gives
(19/3) + (4/3) a6 a0
a7 = − a6 = − = − . (11.96)
(22/3)(22/3) − (1/9) 7 7!
And so on, leading to the following result where the arbitrary constant a0 is replaced
either by arbitrary constant σ1 or by σ2 .
y1 (x) x x2 x3 x4 x5 x6 x7
(1a) : 1 = 1− + − + − + − + ···
σ1 x 3 1! 2! 3! 4! 5! 6! 7!
= exp(−x) .
Similar procedure can now be followed for ν0 ≡ ν2 = (−1/3), and we get
y2 (x) 32 33 34
(1b) : = 1 − 3x + x2 − x3 + x4
σ2 x
−1
3 4 4·7 4 · 7 · 10
35 36
− x5 + x6
4 · 7 · 10 · 13 4 · 7 · 10 · 13 · 16
37
− x7 + · · ·
4 · 7 · 10 · 13 · 16 · 19
For equation III-(2), β = 0 ; γ = −1/2 ; ν = −1/2 ; ρ = −5/2 and the solutions

are
y1 (x) x x2 x3 x4 x5
III − (2a) : 5
= 1+ + + + +
σ1 x 2 9 198 7, 722 463, 320 39, 382, 200
x6 x7
+ + + O(x8 )
4, 489, 570, 800 659, 966, 907, 600
y2 (x) x x2 x3 x4 x5
III − (2b) : = 1 − + − − −
σ2 x−1 5 30 90 360 5400
x6 x7
− − + ···
162, 000 7, 938, 000
For III-(3), β = 3 ; γ = −1 ; ν = 1 ; ρ = −5. Therefore, the two solutions
y1 (x) x √ x2 √
III − (3a) : √ = 1−
(32 + 5 6) + (472 + 183 6)
σ1 x1+ 6 23 460
x 3 √ x 4 √ x5 √
− (1491 + 629 6) + (126 + 47 6) + (−375 + 37 6)
2070 288 720
x6 √ x7 √
+ (5615 − 1731 6) + (−901448 + 333403 6) + · · ·
8640 1512000
y2 (x) x √ x2 √
III − (3b) : √ = 1+
(−32 + 5 6) + (472 − 183 6)
σ2 x1− 6 23 460
x 3 √ x 4 √ x5 √
+ (−1491 + 629 6) + (126 − 47 6) − (375 + 37 6)
2070 288 720
x6 √ x7 √
+ (5615 + 1731 6) − (901448 + 333403 6) + · · ·
8640 1512000
For III-(4), β = 0 ; γ = −1/2 ; ν = 1/2 ; ρ = −5/2 and the solutions are
# $ # $ # $
y1 (x) x x2 x3 x4
III − (4a) : = 1− + − +
5
σ1 x 2 9 9 · 22 9 · 22 · 39 9 · 22 · 39 · 60
# $ # $
x5 x6
− +
9 · 22 · 39 · 60 · 85 9 · 22 · 39 · 60 · 85 · 114
# $
x7
− + ···
9 · 22 · 39 · 60 · 85 · 114 · 147
y2 (x) x x2 x3 x4
III − (4b) : −1
= 1+ + + −
σ2 x 5 5·6 5·6·3 5·6·3·4
# $
x 5 x 6 x7
+ − + + ···
5 · 6 · 3 · 4 · 15 5 · 6 · 3 · 4 · 15 · 30 5 · 6 · 3 · 4 · 15 · 30 · 49
For III-(5), β = −1 ; γ = 5/3 ; ν = −2 ; ρ = −1/3 and the solutions are
y1 (x) x2 x3 x4 x5 x6 x7
III − (5a) : 1 =1+x+ + + + + + + ···
σ1 x 3 2! 3! 4! 5! 6! 7!
= exp(x) .
# $ #$ # $
y2 (x) 32 33 34
III − (5b) : = 1 − 3x − x2 − x3 − x4
σ2 x−1 2 2·5 2·5·8
# $ # $ # $
35 5 36 6 37
− x − x − x7 + ...
2 · 5 · 8 · 11 2 · 5 · 8 · 11 · 14 2 · 5 · 8 · 11 · 14 · 17
For III-(6), β = 1 ; γ = 5/3 ; ν = 2 ; ρ = −1/3 and the solutions are
y1 (x) x2 x3 x4 x5 x6 x7
III − (6a) : 1 =1−x+ − + − + − + ···
σ1 x 3 2! 3! 4! 5! 6! 7!
= exp(−x) .
y2 (x) 32 33 34
III − (6b) : −1
= 1 + 3x − x2 + x3 − x4
σ2 x 2 2·5 2·5·8
# $
35 36 37
+ x5 − x6 + x7 + · · ·
2 · 5 · 8 · 11 2 · 5 · 8 · 11 · 14 2 · 5 · 8 · 11 · 14 · 17
For III-(7), β = 1 ; γ = −3/2 ; ν = 1/2 ; ρ = 1 and the solutions are
y1 (x) x2 x3 x4 x5 x6 x7
III − (7a) : = 1 − x + − + − + − + ···
σ1 x 2 2! 3! 4! 5! 6! 7!
= exp(−x) .
# $ # $ # $
y2 (x) 22 2 23 3 24
III − (7b) : = 1+2x − x + x − x4
σ2 x1/2 1 1·3 1·3·5
# $ # $ # $
25 5 26 6 27
+ x − x + x7 + · · ·
1·3·5·7 1·3·5·7·9 1 · 3 · 5 · 7 · 9 · 11
For III-(8), β = −1 ; γ = −3/2 ; ν = −1/2 ; ρ = 1 and the solutions are
y1 (x) x2 x3 x4 x5 x6 x7
III − (8a) : = 1 + x + + + + + + + ···
σ1 x 2 2! 3! 4! 5! 6! 7!
= exp(x) .
# $ # $ # $
y2 (x) 22 2 23 3 24
III − (8b) : = 1−2x − x − x − x4
σ2 x1/2 1 1·3 1·3·5
# $ # $ # $
25 5 26 6 27
− x − x − x7 + · · ·
1·3·5·7 1·3·5·7·9 1 · 3 · 5 · 7 · 9 · 11
For III-(9), β = −1 ; γ = 1 ; ν = −4/3 ; ρ = −1/9 and the solutions are
y1 (x) x2 x3 x4 x5 x6 x7
III − (9a) : 1 =1+x+ + + + + + + ···
σ1 x 3 2! 3! 4! 5! 6! 7!
= exp(x) .

y2 (x) 31 32 33
III − (9b) : = 1+ x+ x2 + x3
σ2 x − 13 1 1·4 1·4·7
# $
34 35 36
+ x4 + x5 + x6
1 · 4 · 7 · 10 1 · 4 · 7 · 10 · 13 1 · 4 · 7 · 10 · 13 · 16
37
+ x7 + · · ·
1 · 4 · 7 · 10 · 13 · 16 · 19
For III-(10), β = 3 ; γ = −3 ; ν = 1 ; ρ = −4 and the solutions are
y1 (x) x √
III − (10a) : √ = 1− (41 + 22 2)
σ1 x2+2 2 31
x2 √ x3 √
+ (85 + 63 2) − (5, 049 + 3, 457 2)
62 4, 278
x4 √ x5 √
+ (1, 064 + 797 2) − (23, 316 + 11, 485 2)
1, 488 52, 080
x6 √ x7 √
+ (191, 026 − 61, 923 2) + (−264, 230 + 159, 191 2) + · · ·
312, 480 312, 480

y2 (x) x √ x2 √
III − (10b) : √ = 1+ (−41 + 22 2) + (85 − 63 2)
σ2 x2−2 2 31 62
x3 √
+ (−5, 049 + 3, 457 2)
4, 278
x4 √ x5 √
+ (1, 064 − 797 2) + (−23, 316 + 11, 485 2)
1, 488 52, 080
x6 √ x7 √
+ (191, 026 + 61, 923 2) − (264, 230 + 159, 191 2) + · · ·
312, 480 312, 480
Indicial Equation Roots Unequal and Differing by Non-integer

Use (11.85) and set n = 2, 3, 4, 5, 6, 7 in (11.86), and solve the following differential
equations that are of the form (11.78) and belong to category (1). For simplicity, the
index μ has been set equal to zero. The other four indices, β, γ, ν, ρ, are given in
(11.97) (1)–(5) below.
(1) : [β = −3 ; γ = 3 ; ν = 2 ; ρ = −4 ] .
(2) : [β = 2 ; γ = −3 ; ν = 1 ; ρ = −2 ] .
(3) : [β = 3 ; γ = −2 ; ν = −2 ; ρ = −3 ] .
(4) : [β = 1 ; γ = −2 ; ν = −1 ; ρ = −1 ] .
(5) : [β = −1 ; γ = −5 ; ν = −1 ; ρ = −4 ] . (11.97)
11.5 Examples Group IV
Indicial Equation Roots Are Equal

Listed in (11.99) are eight differential equations in the form of (11.98).
βx+γ νx+ρ
y (x) + y (x) + y(x) = 0 . (11.98)
x x2
Equations (A)–(H) are all of category (2): meaning, the two roots, ν1 and ν2 , of their
indicial equation,
ν02 + ν0 (γ − 1) + ρ = 0 ,
are equal—that is ν1 → ν0 and ν2 → ν0 . Note also that these differential equations

have a regular singular point at x = 0. The parameter μ has been set equal to zero in
all of these equations. The other four parameters—namely β, γ, ν, ρ—are listed in
11.5 Examples Group IV 345
a simple format in (11.99) below.
(A) : [β = −3 ; γ = 1 ; ν = −3 ; ρ = 0 ] .
(B) : [β = 4 ; γ = 1 ; ν = −1 ; ρ = 0 ] .
(C) : [β = −2 ; γ = 3 ; ν = −4 ; ρ = 1 ] .
(D) : [β = 1 ; γ = −3 ; ν = 1 ; ρ = 4 ] .
(E) : [β = −1 ; γ = −1 ; ν = −1 ; ρ = 1 ] .
(F) : [β = 1 ; γ = 2 ; ν = 2 ; ρ = 1/4 ] .
(G) : [β = 1 ; γ = −1 ; ν = 1 ; ρ = 1 ] .
(H ) : [β = 1 ; γ = −2 ; ν = 1 ; ρ = 9/4 ] . (11.99)
11.5.1 Solution: (11.99)
As stated above, the two roots of the indicial equation, that is
ν02 + ν0 (γ − 1) + ρ = 0 ,
relating to equations given in (11.99) -(A) to -(H), are equal. Label this double-root
ν0 . Use (11.85). Then set n = 2, 3, 4, 5, 6, 7 in (11.86). And thereby determine the
first solution for each of the eight differential equations listed in (11.99).
The method for working out the first solution of these differential equations is
quite straightforward. Indeed, it is exactly the same method used earlier for the
examples set III. The only difference here is that for (11.99)-(A), similar to Piaggio’s
work 10. the first solution is being written entirely in terms of the variable ν0 . And
when so written, the first solution will also be called the general solution which
is described in detail in the following section. Numerical representation—which is
done by replacing the variable ν0 with its value equal to the relevant double root of the
indicial equation—will be made at the end of the current process. Equation (11.85),
that is similar to (11.100) below, relates a1 to a0 .
β ν0 + ν
a1 = − a0 , (11.100)
2 ν0 + γ
For n ≥ 2, the relevant equation is (11.86) that is similar to (11.101).
β (n − 1 + ν0 ) + ν
an = −an−1 . (11.101)
(n + ν0 )(n + ν0 + γ − 1) + ρ
11.5.2 General Solution: (11.99) -A
Henceforth, in this book, the statement general solution of a given differential equa-
tion will imply a particular form of its first solution that is written as a function of
the variable ν0 . The general solution is determined as follows.
First one considers the indicial equation and determines its roots. In relation to
(11.99)-(A) , where β = −3, γ = 1, ν = −3, ρ = 0, the indicial equation,
ν02 + ν0 (γ − 1) + ρ = ν02 = 0 ,
has two roots, ν0 = ν1 = 0 and ν0 = ν2 = 0. And both the roots are equal to zero.
Second, one identifies the relevant differential equation: it is (11.99)-(A)—written
below as (11.102)—that is obtained from (11.98) by using the given values of the
parameters: β = −3, γ = 1, ν = −3, ρ = 0.
1 − 3x 3
y (x) + y (x) − y(x) = 0 (11.102)
x x
Third, one uses (11.100).
β ν0 + ν 1 + ν0
a1 = − a0 = 3 a0 . (11.103)
2 ν0 + γ 1 + 2 ν0
Fourth, for n ≥ 2, one writes (11.101) as
β (n − 1 + ν0 ) + ν
an = −an−1
(n + ν0 )(n + ν0 + γ − 1) + ρ
3
= an−1 . (11.104)
(n + ν0 )
As such
3 32 (1 + ν0 )
a2 = a1 = a0 ,
(2 + ν0 ) (2 + ν0 )(1 + 2ν0 )
3 33 (1 + ν0 )
a3 = a2 = a0 ,
(3 + ν0 ) (2 + ν0 )(3 + ν0 )(1 + 2ν0 )
3 34 (1 + ν0 )
a4 = a3 = a0 ,
(4 + ν0 ) (2 + ν0 )(3 + ν0 )(4 + ν0 )(1 + 2ν0 )
3 35 (1 + ν0 )
a5 = a4 = a0 ,
(5 + ν0 ) (2 + ν0 )(3 + ν0 )(4 + ν0 )(5 + ν0 )(1 + 2ν0 )
36 (1 + ν0 )
a6 = a0 ,
(2 + ν0 )(3 + ν0 )(4 + ν0 )(5 + ν0 )(6 + ν0 )(1 + 2ν0 )
11.5 Examples Group IV 347
37 (1 + ν0 )
a7 = a0 .
(2 + ν0 )(3 + ν0 )(4 + ν0 )(5 + ν0 )(6 + ν0 )(7 + ν0 )(1 + 2ν0 )
(11.105)
The first solution, YA;1 (x, ν0 ), is all contained in (11.103) and (11.105). As stated
above, this solution—meaning (11.106)—will henceforth also be called the general
solution, G A (x, ν0 ), of (11.99)-(A) or equivalently of (11.102).
G A (x, ν0 )
= YA;1 (x, ν0 )
σ1
3 (1 + ν0 ) x 32 (1 + ν0 ) x2 33 (1 + ν0 ) x3
= xν0 [1 + + +
(1 + 2 ν0 ) (2 + ν0 )(1 + 2ν0 ) (2 + ν0 )(3 + ν0 )(1 + 2ν0 )
34 (1 + ν0 ) x4
+
(2 + ν0 )(3 + ν0 )(4 + ν0 )(1 + 2ν0 )
35 (1 + ν0 ) x5
+
(2 + ν0 )(3 + ν0 )(4 + ν0 )(5 + ν0 )(1 + 2ν0 )
36 (1 + ν0 ) x6
+
(2 + ν0 )(3 + ν0 )(4 + ν0 )(5 + ν0 )(6 + ν0 )(1 + 2ν0 )
37 (1 + ν0 ) x7
+ ] + O(x8 )
(2 + ν0 )(3 + ν0 )(4 + ν0 )(5 + ν0 )(6 + ν0 )(7 + ν0 )(1 + 2ν0 )
(11.106)
The double root for (11.99)-(A), i.e. (11.102), is ν0 = 0. Therefore, the first so-
lution, YA;1 (x, ν0 ), for (11.99)-(A) is immediately available in numerical format
YA;1 (x, ν0 = 0), by setting ν0 = 0 in its general solution (11.106). [See, for instance,
(11.107).]
YA;1 (x, ν0 = 0) YA;1 (x, 0) G A (x, ν0 = 0)

(A) : ≡ =
σ1 σ1 σ1
2 3 4 5
9 x 9 x 27 x 81 x 81 x6 243 x7
= x0 [1 + 3 x + + + + + + + ···]
2 2 8 40 80 560
= x0 exp(3x) = exp(3x). (11.107)
For YA;1 (x, 0), rather than σ0 , it was convenient to use σ1 as the arbitrary constant.
Similarly, for the second solution YA;2 (x, 0)—to be studied later—σ2 would be the
relevant arbitrary constant.
11.6 First Solution of (11.99)-(B)→(E)
First solution of differential equations (B), (C), (D), (E), (F), (G), and (H ) can be
found in similar fashion and the results are as follows.
For differential equation (B), β = 4, γ = 1, ν = −1 and ρ = 0. The double root

of the indicial equation is again ν0 = 0 and the first solution is
YB;1 (x, 0) 3 x2 7 x3 77 x4 77 x5 1, 463 x6

(B) : = 1 + x − + − + −
σ1 x 0 4 12 192 320 11, 520
4, 807 x7
+ + ··· (11.108)
80, 640
For differential equation (C), β = −2, γ = 3, ν = −4 and ρ = 1, the double root

of the indicial equation is ν0 = −1 and its first solution is
YC;1 (x, −1) 4 x3 2 x4 4 x5 4 x6 8 x7

(C) : = 1 + 2 x + 2 x 2
+ + + + + + ···
σ1 x−1 3 3 15 45 315
= exp(2x) . (11.109)
For differential equation (D), β = 1, γ = −3, ν = 1 and ρ = 4, the double root

of the indicial equation is ν0 = 2 and its first solution is
YD;1 (x, 2) 5 x3 5 x4 7 x5 7 x6 x7
(D) : = 1 − 3x + 3 x2 − + − + − + ···
σ1 x 2 3 8 40 180 140
(11.110)
For differential equation (E), β = −1, γ = −1, ν = −1 and ρ = 1, the double

root of the indicial equation is ν0 = 1 and its first solution is
YE;1 (x, 1) 3 x2 2 x3 5 x4 x5 7 x6 x7
(E) : = 1 + 2x + + + + + + + ···
σ1 x 2 3 24 20 720 630
(11.111)
For differential equation (F), β = 1, γ = 2, ν = 2 and ρ = 1/4, the double root

of the indicial equation is ν0 = −1/2 and its first solution is

YF;1 (x, −1/2) 3x 15 x2 35 x3 105 x4 231x5 1, 001 x6
(F) : 1
=1− + − + − +
σ1 x− 2 2 16 96 1, 024 10, 240 245, 760
143x7
− + ··· (11.112)
229, 376
For differential equation (G), β = 1, γ = −1, ν = 1 and ρ = 1, the double root

of the indicial equation is ν0 = 1. Later on, both its first and second solutions will
be worked out and details presented—as, for instance, in (11.123) and (11.132). In
the meantime the first solution is given below.
11.6 First Solution of (11.99)-(B)→(E) 349
YG;1 (x, 1) 3 x2 2 x3 5 x4 x5 7 x6 x7
(G) : = 1 − 2x + − + − + − + ···
σ1 x 2 3 24 20 720 630
= (1 − x) exp(−x) (11.113)
Finally, for differential equation (H ), β = 1, γ = −2, ν = 1 and ρ = 9/4, the

double root of the indicial equation is ν0 = 3/2 and its first solution is
YH ;1 (x, 3/2) 5x 35 x2 35 x3 385 x4 1, 001 x5

(H ) : 3 = 1− + − + −
σ1 x 2 2 16 32 1, 024 10, 240
1, 001 x6 2, 431 x7
+ − + ··· . (11.114)
49, 152 688, 128
11.7 Methodology For Second Solution
Equation (11.99)-(A)–(H) are of category (2): Meaning their indicial equations have
double roots. The methodology to be used for working out their second solution will
involve two different methods: one inspired by a method used by Piaggio10. and
the other based on what is known as the ‘method of order reduction.’ The relevant
details are provided in the following. For convenience, these methods are termed as
the ‘second methods.’
11.7.1 Piaggio-Like Solution
Piaggio, while studying a particular differential equation whose indicial equation

roots are equal, came up with an idea about the second solution. Assuming that his
idea holds true for the differential equation being studied here, the two solutions of
a category (2) differential equation should have the following characteristics.
The first solution is the general solution itself while the second solution is equal
to the differential with respect to ν0 of the general solution. At the end of the process,
one needs to convert the first- and the second solution into numerical format. This is
done by replacing the symbol ν0 by its actual value: which is the relevant double root
of the indicial equation. This procedure will become clear as the process unfolds.
Consider differential equation (11.99)-(A), its general solution—see (11.106)—
G A (x, ν0 ), and proceed as follows.
YA;1 (x, ν0 ) = G A (x, ν0 ), (11.115)

and
d YA;1 (x, ν0 ) d G A (x, ν0 )
YA;2 (x, ν0 ) = = . (11.116)
d ν0 d ν0
Because YA;1 (x, ν0 ) includes the term xν0 , the differential of xν0 needs to be worked
out. To that end, define a symbol X (ν0 , x).
X (ν0 , x) = xν0 . (11.117)
And proceed as follows. Take logarithm of both sides.
log X (ν0 , x) = log(xν0 ) = ν0 log(x) , (11.118)
and differentiate with respect to ν0 .
1 dx(ν0 , x)
· = log(x) , (11.119)
X (ν0 , x) d ν0
Note: Any remaining terms that would arise if x itself depended on ν0 are not included
in (11.119). Now rewrite the above as
dx(ν0 , x)
= X (ν0 , x) log(x) . (11.120)
d ν0
and replace X (ν0 , x) by xν0 .
dxν0
= xν0 log(x) . (11.121)
d ν0
Accordingly, the second solution is
d YA;1 (x, ν0 ) d G A (x, ν0 )

YA;2 (x, ν0 ) ≡ =
d ν0 d ν0
3x 3 x2 6 x2
= YA;1 (x, ν0 ) · log(x) − σ1 xν0 + +
(1 + 2 ν0 ) 2 (2 + ν0 ) 2 (1 + 2ν0 )2

11 + 22ν0 + 17ν0 2 + 4ν0 3
− σ1 xν0 27 x3
(2 + ν0 )2 (3 + ν0 )2 (1 + 2ν0 )2
(25 + 61ν0 + 59ν0 2 + 23ν0 3 + 3ν0 4 )
− σ1 xν0 162 x4
(2 + ν0 )2 (3 + ν0 )2 (4 + ν0 )2 (1 + 2ν0 )2
(274 + 758ν0 + 847ν0 2 + 428ν0 3 + 97ν0 4 ) + 8ν0 5 ) 5
− σ 1 x ν0 243 x
(2 + ν0 )2 (3 + ν0 )2 (4 + ν0 )2 (5 + ν0 )2 (1 + 2ν0 )2
− σ 1 x ν0 [etc., etc., · · · ] (11.122)
11.7 Methodology For Second Solution 351
Equation (11.122) contains all the essentials of the second solution. However, in
order to convert it into proper form, one needs to do two simple things.
(i) : Replace the arbitrary constant σ1 in (11.122) by an arbitrary constant σ2
everywhere.
(ii) : Knowing that the double root of the indicial equation for (11.99)-(A) is
ν0 = 0, everywhere in (11.122), replace the symbol ν0 with its value 0.
Once these changes have been executed, the numerical form of the second solution,
YA;2 (x, 0), of differential equation (11.99)-(A) can be written as
YA;2 (x, 0) 27 2 33 3
= exp(3 x) · log(x) − 3 x − x − x
σ2 Piaggio 4 4
225 3, 699 3, 969
− x4 − x5 − x6
32 800 1600
88, 209 554, 769
− x7 − x8 − O(x9 ) . (11.123)
78, 400 1, 254, 400
Note: In going from (11.122) to (11.123), we have used the result, G A (x, ν0 = 0) =
YA;1 (x, 0) = σ2 exp(3x), for the first solution that was given in (11.107). Note also
that the arbitrary constant σ1 has been replaced by arbitrary constant σ2 .
11.7.2 Method of Order Reduction
If a nonzero first solution, y1 (x), of a second-order homogeneous linear ordinary

differential equation is already known, a linearly independent second solution can be
determined by a procedure that was studied in detail in a previous chapter in relation
to (7.31)–(7.45). That procedure consists in representing the second solution y2 (x)
as a product of the first solution y1 (x) and an unknown new function f (x).
y2 (x) = f (x) y1 (x) . (11.124)
In this regard, (7.31)–(7.45) can be used without alteration as long as the old notation
is aligned with the current notation. To that purpose, consider first the differential
equation (7.31) presented below as (11.125)

a2 (x) D2 + a1 (x) D + a0 (x) u(x) = 0 , (11.125)
and compare it to the current differential equation (11.126).
1 − 3x 3
y (x) + y (x) − y(x) = 0 . (11.126)
x x
By comparing (11.125) and (11.126), we get
1 − 3x 3
a2 (x) = 1 ; a1 (x) = ; a0 (x) = − (11.127)
x x
According to (7.45), y2 (x) is related to y1 (x) as in (11.128).

&
% exp − a1 (x)
dx
y2 (x) a2 (x)
= f (x) = σ3 · dx . (11.128)
y1 (x) [y1 (x)]2
Equation (11.128) involves two integrations. First the integral

% %
a1 (x) 1 − 3x
− dx = − dx = − log(x) + 3 x . (11.129)
a2 (x) x
As a result, (11.128) can be written as

%
exp(3 x)
y2 (x) = σ3 y1 (x) · dx. (11.130)
x [y1 (x)]2
Next, using the fact that y1 (x) = const. exp(3 x), we have
%
exp(−3x)
y2 (x) = σ4 exp(3x) dx . (11.131)
x
The integral in (11.131) is readily done.

%
exp(−3x)
dx
x
%
1 9 9 27 81 81 243 6
= − 3 + x − x2 + x3 − x4 + x5 − x ... dx
x 2 2 8 40 80 560
9 3 27 81 5 27 6 243 7
= log(x) − 3x + x2 − x3 + x4 − x + x − x ...
4 2 32 200 160 3920
As a result, (11.131) becomes
y2 (x)
= exp(3x) log(x)
σ4
9 3 27 81 5
+ exp(3x) −3x + x2 − x3 + x4 − x + O(x6 )
4 2 32 200
Finally, we need the power expansion of exp(3x).
9 9 27 81 81 243 7 729 8
exp(3x) = 1 + 3x + x2 + x3 + x4 + x5 + x6 + x + x
2 2 8 40 80 560 4, 480
This leads to
y2 (x)
= exp(3x) log(x)
σ4
9 9 27 9 3 27
+ 1 + 3x + x2 + x3 + x4 + etc. . −3x + x2 − x3 + x4 − etc.
2 2 8 4 2 32
As a result, the numerical representation of the second solution of differential equa-

tion (11.99)-(A), calculated by the method of order reduction, is the following.
YA;2 (x, 0)
σ4 ord.reduction
27 2 33 3
= exp(3 x) · log(x) − 3 x − x − x
4 4
225 3, 699 3, 969
− x4 − x5 − x6
32 800 1600
88, 209 554, 769
− x7 − x8
78, 400 1, 254, 400
192, 483 597, 861
−− x9 − x10 , etc. (11.132)
1, 254, 400 12, 544, 000

YA;2 (x,0) YA;2 (x,0)
Note: σ4
in (11.132) is the same as σ4
in (11.123). As
ord.reduction Piaggio
usual, σ0 , σ1 , σ2 , etc., are arbitrary constants.
11.7.3 Complete Solution of (11.99)-(A)
Complete numerical solution of (11.99)-(A), YA;Complete (x, 0), consists of YA;1 (x, 0)
recorded in (11.107) and YA;2 (x, 0) in (11.132). As such, (11.133)—which is the sum
of these two—represents the complete solution of (11.99)-(A).
YA;0 (x, 0) = YA;1 (x, 0) + YA;2 (x, 0)

= σ1 exp(3x)
27 2 33 3
+ σ2 [exp(3 x) · log(x) − 3 x − x − x
4 4
225 3, 699 3, 969
− x4 − x5 − x6
32 800 1600
88, 209 554, 769

− x7 − x8
78, 400 1, 254, 400
192, 483 597, 861
−− x9 − x10 ] ... . (11.133)
1, 254, 400 12, 544, 000
11.7.4 General Solution of (11.99)-(G)
For (11.99)-(G) the prescribed values of the variables are: β = 1, γ = −1, ν = 1

and ρ = 1. As such the form of differential equation (11.98) that is being considered
here is (11.134).
x−1 x+1
y (x) + y (x) + y(x) = 0 . (11.134)
x x2
The relevant indicial equation,
ν02 + ν0 (γ − 1) + ρ = ν02 − 2ν0 + 1 = 0 ,
has a double root: ν1 = ν0 = 1 and ν2 = ν0 = 1. As before, we work directly with

the variable ν0 and do not specify its numerical value. As such, (11.85), that relates
a1 to a0 , is written as
ν + β ν0 1 + ν0
a1 = −a0 = − a0 . (11.135)
γ + 2 ν0 2 ν0 − 1
[Also see (11.100).] To calculate a2 and higher-order an , rewrite (11.90) [Also see
(11.101)].
β (n − 1 + ν0 ) + ν
an = −an−1
(n + ν0 )(n + ν0 + γ − 1) + ρ
(n − 1 + ν0 ) + 1
= −an−1
(n + ν0 )(n + ν0 − 1 − 1) + 1
(n + ν0 )
= −an−1 (11.136)
(n + ν0 − 1)2
For n = 2, (11.135) and (11.136) give
(2 + ν0 )
a2 = a0 . (11.137)
(ν0 + 1)(2 ν0 − 1)
For n = 3, (11.135) and (11.136) give

(3 + ν0 )
a3 = − a0 . (11.138)
(ν0 + 2)(ν0 + 1)(2 ν0 − 1)
For n = 4, (11.135) and (11.136) give
(4 + ν0 )
a4 = a0 . (11.139)
(ν0 + 3)(ν0 + 2)(ν0 + 1)(2 ν0 − 1)
For n = 5, (11.135) and (11.136) give
(5 + ν0 )
a5 = − a0 . (11.140)
(ν0 + 4)(ν0 + 3)(ν0 + 2)(ν0 + 1)(2 ν0 − 1)
For n = 6, 7, 8, (11.135) and (11.136) give
(6 + ν0 )
a6 = a0
(ν0 + 5)(ν0 + 4)(ν0 + 3)(ν0 + 2)(ν0 + 1)(2 ν0 − 1)
(7 + ν0 )
a7 = − a0
(ν0 + 6)(ν0 + 5)(ν0 + 4)(ν0 + 3)(ν0 + 2)(ν0 + 1)(2 ν0 − 1)
(8 + ν0 )
a8 = a0 .
(ν0 + 7)(ν0 + 6)(ν0 + 5)(ν0 + 4)(ν0 + 3)(ν0 + 2)(ν0 + 1)(2 ν0 − 1)
(11.141)
The above results—that is (11.135)–(11.141)—are listed below.
G G (x, ν0 )
σ 0 ν0 x ν 0
1 + ν0 (2 + ν0 )
= 1− x+ x2
2 ν0 − 1 (ν0 + 1)(2 ν0 − 1)
(3 + ν0 ) (4 + ν0 )
− x3 + x4
(ν0 + 2)(ν0 + 1)(2 ν0 − 1) (ν0 + 3)(ν0 + 2)(ν0 + 1)(2 ν0 − 1)
(5 + ν0 )
− x5
(ν0 + 4)(ν0 + 3)(ν0 + 2)(ν0 + 1)(2 ν0 − 1)
(6 + ν0 )
+ x6
(ν0 + 5)(ν0 + 4)(ν0 + 3)(ν0 + 2)(ν0 + 1)(2 ν0 − 1)
(7 + ν0 )
− x7
(ν0 + 6)(ν0 + 5)(ν0 + 4)(ν0 + 3)(ν0 + 2)(ν0 + 1)(2 ν0 − 1)
(8 + ν0 )
+ x8
(ν0 + 7)(ν0 + 6)(ν0 + 5)(ν0 + 4)(ν0 + 3)(ν0 + 2)(ν0 + 1)(2 ν0 − 1)
− O(x9 ) (11.142)
G G (x, ν0 ) in (11.142) is the general solution of (11.99)-(G).
11.7.5 First Solution of (11.99)-(G)
For (11.99)-(G), β = 1, γ = −1, ν = 1 and ρ = 1 and the relevant differential equa-

tion is (11.134). The double root of the indicial equation is ν0 = 1. As stated before,
the first solution, YG;1 (x, ν0 ), is the general solution, G G (x, ν0 ), itself while the sec-
ond solution, YG;2 (x, ν0 ), is the differential with respect to ν0 of the general solution.
And at the end of the process, one converts the first and the second solution into
numerical format. This is done by replacing the symbol ν0 by its actual value: which
is the double root of the relevant indicial equation. The notation to be used is the
following.
For unspecified value of ν0 the first and the second solutions of (11.99)-(G) are
denoted as YG;1 (x, ν0 ) and YG;2 (x, ν0 ). And furthermore, as stated earlier,
YG;1 (x, ν0 ) = G G (x, ν0 ) , (11.143)
and
d YG;1 (x, ν0 ) d G G (x, ν0 )

YG;2 (x, ν0 ) = = . (11.144)
d ν0 d ν0
For specific value of ν0 equal to the double root of the indicial equation—which
for the present case is ν0 = 1—the first and the second solutions of (11.99)-(G) are
to be denoted as
G G (x, ν0 = 1) = YG;1 (x, ν0 = 1) ≡ YG;1 (x, 1) , (11.145)
and
d YG;1 (x, ν0 )
YG;2 (x, ν0 = 1) = YG;2 (x, 1) = . (11.146)
d ν0 ν0 =1
Setting ν0 = 1 everywhere in the general solution G G (x, ν0 ), that was given in

(11.142), leads to the following result for the first solution, YG;1 (x, 1), of (11.99)-(G).
[Note: Compare (11.113).]
YG;1 (x, 1) G G (x, ν0 = 1)

(G) : = = (1 − x) exp(−x)
σ1 x σ1 x
3 x2 2 x3 5 x4 x5 7 x6 x7
= 1− 2x + − + − + − + ··· . (11.147)
2 3 24 20 720 630
11.7.6 Second Solution of (11.99)-(G)
To save space, only the Piaggio-like second solution is being presented.
11.7.7 Piaggio-Like Second Solution
As stated above and expressed in (11.144), for unspecified values of ν0 , the second
solution, YG;2 (x, ν0 ), of (11.99)-(G) is the differential, with respect to the variable
ν0 , of its general solution given in (11.142). That is
d G G (x, ν0 )
YG;2 (x, ν0 ) =
d ν0
(1 + 2ν0 − 2ν02 )
= G G (x, ν0 ) · log(x) + σ1 xν0 1 + x
(2 ν0 − 1)2
2 + 2ν0 + 3ν0 2
− σ 1 x ν0 x2
(2ν02 + ν0 − 1)2
(3 + 2ν0 + 7ν0 2 + 6ν0 3 + ν0 4 ) 3
+ σ 1 x ν0 2 x
(2ν03 + 5ν02 + ν0 − 2)2
24 + 12ν0 + 63ν0 2 + 88ν0 3 + 35ν04 + 4ν05
− σ 1 x ν0 x4
(2ν04 + 11ν03 + 16ν02 + ν0 − 6)2
(2ν06 + 26ν05 + 115ν04 + 200ν03 + 109ν02 + 16ν0 + 40 5
+ σ 1 x ν0 3 x
[(ν0 + 4)(ν0 + 3)(ν0 + 2)(ν0 + 1)(2ν0 − 1)]2
− O(xν0 x6 ) (11.148)
(i) : Replace the arbitrary constant σ1 in (11.148) by an arbitrary constant σ2
everywhere.
(ii) : Knowing that the double root of the indicial equation for (11.99)-(G) is
ν0 = 1, everywhere in (11.148), replace the symbol ν0 with its value 1.
Once these changes have been introduced the second solution,
YG;2 (x, ν0 = 1) → YG;2 (x), of (11.99)-(G) can be written in its final form.
YG ; 2 (x, ν0 = 1) YG ; 2 (x)
=
σ2 Piaggio σ2 Piaggio
7 3 19 4 113 5
= x(1 − x) exp(−x) log(x) + x + x2 − x + x − x
4 18 288
127
+ x6 − O(x7 ) (11.149)
1, 200
Note: In going from (11.148) to (11.149), we have used the result, G G (x, ν0 = 1) =
YG;1 (x, ν0 = 1) = YG;1 (x) = σ1 x(1 − x) exp(−x) for the first solution that was
given in (11.147). Of course, the arbitrary constant σ1 there has been changed to σ2
here.
Exercise for the Reader: (11.99)-(B)
Follow the above procedure and find the second solution for differential equa-
tion (11.99)-(B).
11.8 Examples Group V
Consider the differential equation.
βx+γ νx+ρ
y (x) + y (x) + y(x) = 0 . (11.150)
x x2
The four parameters β, γ, ν, ρ referred to in (11.150) are provided in (11.151).
(1) : [β = 1 ; γ = −2 ; ν = 1 ; ρ = 2 ] .
(2) : [β = 1 ; γ = −3 ; ν = 1 ; ρ = 3 ] .
(3) : [β = 1 ; γ = −4 ; ν = 1 ; ρ = 4 ] .
(4) : [β = 1 ; γ = −5 ; ν = 1 ; ρ = 5 ] .
(5) : [β = 1 ; γ = −6 ; ν = 1 ; ρ = 6 ] . (11.151)
Use (11.85) ; set n = 2, 3, 4, 5, 6, 7 in (11.86) and remember that the first solution
of these differential equations makes use of the larger of the two roots of the indicial
equation: that is ν0 = ν1 . The smaller root, ν0 = ν2 , of the indicial equation is used
for the second solution. Because of space constraints, provide detailed information
about the general solution, first solution, and two different techniques for working
out the second solution, only for (11.151)-(4).
11.8.1 General Solution of (11.151)-(4)
For (11.151)-(4)—that is of the form of (11.150)—the prescribed values of the vari-

ables are: β = 1, γ = −5, ν = 1 and ρ = 5. Accordingly, the relevant differential
equation to consider is
x−5 x+5
y (x) + y (x) + y(x) = 0 . (11.152)
x x2
11.8 Examples Group V 359
The two roots of its indicial equation,
ν02 + ν0 (γ − 1) + ρ = ν02 − 6ν0 + 5 = 0 ,
are ν0 = 5 and ν0 = 1. Work directly with the variable ν0 and not specify either of
its two roots. As such, (11.85), that relates a1 to a0 , is written as
ν + β ν0 1 + ν0
a1 = −a0 = − a0 . (11.153)
γ + 2 ν0 2 ν0 − 5
[Note: Compare (11.100).] To calculate a2 and higher-order an , rewrite (11.90) [Note:

See (11.100).]
β (n − 1 + ν0 ) + ν
an = −an−1
(n + ν0 )(n + ν0 + γ − 1) + ρ
(n − 1 + ν0 ) + 1
= −an−1
(n + ν0 )(n + ν0 − 5 − 1) + 5
(n + ν0 )
= −an−1 . (11.154)
(n + ν0 − 5)(n + ν0 − 1)
For n = 2, (11.153) and (11.154) give
(2 + ν0 )
a2 = a0 . (11.155)
(ν0 − 3)(2 ν0 − 5)
For n = 3, (11.153) and (11.154) give
(3 + ν0 )
a3 = − a0 . (11.156)
(ν0 − 3)(ν0 − 2)(2 ν0 − 5)
For n = 4, (11.153) and (11.154) give
(4 + ν0 )
a4 = a0 . (11.157)
(ν0 − 3)(ν0 − 2)(ν0 − 1)(2 ν0 − 5)
For n = 5, (11.153) and (11.154) give
(5 + ν0 )
a5 = − a0 . (11.158)
(ν0 − 3)(ν0 − 2)(ν0 − 1)(ν0 )(2 ν0 − 5)
For n = 6, 7, 8, (11.153) and (11.154) give
(6 + ν0 )
a6 = a0
(ν0 − 3)(ν0 − 2)(ν0 − 1)(ν0 )(ν0 + 1)(2 ν0 − 5)
(7 + ν0 )
a7 = − a0
(ν0 − 3)(ν0 − 2)(ν0 − 1)(ν0 )(ν0 + 1)(ν0 + 2)(2 ν0 − 5)
(8 + ν0 )
a8 = a0 .
(ν0 − 3)(ν0 − 2)(ν0 − 1)(ν0 )(ν0 + 1)(ν0 + 2)(ν0 + 3)(2 ν0 − 5)
(11.159)
These results are listed below.
G(x, ν0 )
σ 0 x ν0
1 + ν0 (2 + ν0 )
= 1− x+ x2
2 ν0 − 5 (ν0 − 3)(2 ν0 − 5)
(3 + ν0 ) (4 + ν0 )
− x3 + x4
(ν0 − 3)(ν0 − 2)(2 ν0 − 5) (ν0 − 3)(ν0 − 2)(ν0 − 1)(2 ν0 − 5)
(5 + ν0 )
− x5
(ν0 − 3)(ν0 − 2)(ν0 − 1)(ν0 )(2 ν0 − 5)
(6 + ν0 )
+ x6
(ν0 − 3)(ν0 − 2)(ν0 − 1)(ν0 )(ν0 + 1)(2 ν0 − 5)
(7 + ν0 )
− x7
(ν0 − 3)(ν0 − 2)(ν0 − 1)(ν0 )(ν0 + 1)(ν0 + 2)(2 ν0 − 5)
(8 + ν0 )
+ x8
(ν0 − 3)(ν0 − 2)(ν0 − 1)(ν0 )(ν0 + 1)(ν0 + 2)(ν0 + 3)(2 ν0 − 5)
− O(x9 ) (11.160)
G(x, ν0 ) in (11.160) is the general solution of (11.152) that represents (11.151)-(4).
11.8.2 First Solution of (11.151)-(1)–(5)
Consider differential equation (11.150) and set the values of β, γ, ν, ρ as provided in

(11.151)-(1)–(5). For differential equation (11.151)-(1), β = 1, γ = −2, ν = 1 and
ρ = 2, and the two roots of the indicial equation,
ν02 + ν0 (γ − 1) + ρ = 0 = ν02 − 3ν0 + 2 ,
are ν0 = 2 and ν0 = 1. Accordingly, setting ν0 = 2 everywhere in its general solution

gives the first solution, y1;1 (x, ν0 = 2), for (11.151)-(1). [Note, to save space, neither
the general solution nor the first solution for (11.151)-(1), (2), (3) and (5) are recorded
11.8 Examples Group V 361
here.]
y1;1 (x, ν0 = 2) 3x 5 x3 x4 7 x5 x6 x7
(1) : = 1 − + x 2
− + − + − + ···
σ1 x 2 2 12 8 240 180 1, 120
x
= 1− exp(−x). (11.161)
2
For differential equation (11.151)-(2), β = 1, γ = −3, ν = 1 and ρ = 3, and the
two roots of the indicial equation are ν0 = 3 and ν0 = 1. The following result obtains
for its first solution, y1;2 (x, ν0 = 3).
y1;2 (x, ν0 = 3) 4x 5 x2 x3 7 x4 x5 x6 x7
(2) : =1− + − + − + − + ···
σ1 x 3 3 6 3 72 45 240 1, 512
x
= 1− exp(−x) . (11.162)
3

two roots of the indicial equation are ν0 = 4 and ν0 = 1. We get
y1;3 (x, ν0 = 4) 5x 3 x2 7 x3 x4 3 x5 x6 11 x7
(3) : =1− + − + − + − + ···
σ1 x 4 4 4 24 12 160 288 20, 160
x
= 1− exp(−x) . (11.163)
4

two roots of the indicial equation are ν0 = 5 and ν0 = 1. Therefore,
y1;4 (x, ν0 = 5) 6x 7 x2 4 x3 3 x4 x5 11 x6 x7
(4) : =1− + − + − + − + ···
σ1 x 5 5 10 15 40 60 3, 600 2, 100
x
= 1− exp(−x) . (11.164)
5
And finally for differential equation (11.151)-(5), we have β = 1, γ = −6, ν = 1

and ρ = 6, and the two roots of the indicial equation are ν0 = 6 and ν0 = 1. For root
ν0 = 6, the result for the first solution is
y1;5 (x, ν0 = 6) 7x 2 x2 x3 5 x4 11 x5 x6 13 x7
(5) : =1− + − + − + − + ···
σ1 x 6 6 3 4 72 720 360 30, 240
x
= 1− exp(−x) . (11.165)
6
11.9 Equation (11.151)-(4)
11.9.1 Second Solutions
Use the prescribed values of the variables noted in (11.151)-(4). That is β = 1, γ =

−5, ν = 1 and ρ = 5. Insert them into differential equation (11.150). This leads
to (11.166) which is the relevant differential equation for (11.151)-(4).
x−5 x+5
y (x) + y (x) + y(x) = 0 . (11.166)
x x2
The two roots of its indicial equation,
ν02 + ν0 (γ − 1) + ρ = ν02 − 6ν0 + 5 = 0 ,
are ν0 = 5 and ν0 = 1.
The second solution uses the smaller of the two roots namely ν0 = 1. A cursory
look at (11.151)-(4)’s general solution (11.160) shows how it is replete with infinities.
Infinities arise from the term (ν0 − 1) in the denominator for the reason that this term
goes to zero when ν0 = 1. To obviate the occurrence of these infinities, one proceeds
as follows.
11.9.2 Piaggio’s Solution
Replace the arbitrary constant σ0 in the general solution (11.160) by σpia;4 (ν0 − 1)
where σpia;4 is also an arbitrary constant. Doing this transforms G(x, ν0 ) into
G pia (x, ν0 ). More importantly, it gets rid everywhere of the pesky denominator
(ν0 − 1). As a result, the general solution (11.160) changes into
' (
G pia (x, ν0 ) 1 + ν0
= (ν 0 − 1) 1 − x
σpia;4 xν0 2 ν0 − 5
' (
(2 + ν0 ) (3 + ν0 )
+ (ν0 − 1) x −
2
x 3
(ν0 − 3)(2 ν0 − 5) (ν0 − 3)(ν0 − 2)(2 ν0 − 5)
(4 + ν0 ) (5 + ν0 )
+ x4 − x5
(ν0 − 3)(ν0 − 2)(2 ν0 − 5) (ν0 − 3)(ν0 − 2)(ν0 )(2 ν0 − 5)
(6 + ν0 )
+ x6
(ν0 − 3)(ν0 − 2)(ν0 )(ν0 + 1)(2 ν0 − 5)
(7 + ν0 )
− x7
(ν0 − 3)(ν0 − 2)(ν0 )(ν0 + 1)(ν0 + 2)(2 ν0 − 5)
11.9 Equation (11.151)-(4) 363
(8 + ν0 )
+ x8
(ν0 − 3)(ν0 − 2)(ν0 )(ν0 + 1)(ν0 + 2)(ν0 + 3)(2 ν0 − 5)
− O(x9 ) (11.167)
The next thing to do is to differentiate G pia (x, ν0 ) with respect to ν0 . And call the
result y2;4 (x, ν0 ).
d G pia (x, ν0 )
= y2;4 (x, ν0 )
d ν0
= G pia (x, ν0 ) log(x)
ν02 − 5ν0 + 1 27 10
+ σpia;4 xν0 · [1 − 2 x+ − x2
(2ν0 − 5)2 (2ν0 − 5)2 (ν0 − 3)2
66 12 5
− − − x3
(2ν0 − 5)2 (ν0 − 3)2 (ν0 − 2)2
4ν03 + 9ν02 − 120ν0 + 178
− x4
(2ν03 − 15ν02 + 37ν0 − 30)2
3ν04 + 5ν03 − 94ν02 + 185ν0 − 75 5
+2 x
ν 2 (2ν03 − 15ν02 + 37ν0 − 30)2
8ν05 + 21ν04 − 268ν03 + 403ν02 + 84ν0 − 180 6
− x
ν02 (2ν04 − 13ν03 + 22ν02 + 7ν0 − 30)
10ν06 + 48ν05 − 327ν04 − 10ν03 + 1055ν02 − 224ν0 − 420 7
+ x
ν02 (2ν05 − 9ν04 − 4ν03 + 51ν02 − 16ν0 − 60)2
− O(x8 )] (11.168)
Set ν0 = 1 in both (11.167) and (11.168). Equation (11.167) gives
G pia (x, ν0 = 1)
σpia;4
5x5 6x 7x2 8x3 9x4
=− 1− + − + − O(x5 )
6 5 2!5 3!5 4!5
5 5 x
=− x 1− exp(−x) , (11.169)
6 5
and (11.168) yields
y2;4 (x, ν0 = 1)
σpia;4
G pia (x, ν0 = 1) log(x) 2 2 1 3 2 4
= +x+ x + x + x
σpia;4 3 2 3
− O(x5 ). (11.170)
Combining (11.169) and (11.170) leads to Piaggio’s suggested second solution of

(11.151)-(4).
y2;4 (x, ν0 = 1) y2;4 (x, ν0 = 1)

≡
σpia;4 σpia;4 piaggio
5 5 x 2 2 1 3 2 4
=− x 1− exp(−x) . log(x) + x + x + x + x
6 5 3 2 3
− O(x5 ). (11.171)
11.10 Solution by Method of Order Reduction
As explained earlier, given the availability of a nonzero solution, y1 (x), of a second-

order homogeneous linear ordinary differential equation a linearly independent sec-
ond solution can be determined by a procedure known as order reduction. [Compare
(7.31)–(7.45).] This procedure consists in representing the second solution,
y2 (x),
as a product of the first solution, which currently is y1 (x) = σ1 x5 1 − 5x exp(−x),
and an unknown new function f (x).
y2 (x) = f (x) y1 (x) . (11.172)
For a differential equation

a2 (x) D2 + a1 (x) D + a0 (x) u(x) = 0 (11.173)
the second solution, y2 (x), is given by the relationship

&
% exp − a1 (x) dx
a2 (x)
y2 (x) = y1 (x) f (x) = σ0 y1 (x) · dx . (11.174)
[y1 (x)]2
Note, σ0 is an arbitrary constant. By comparing (11.173) and the current differential

equation,
x−5 x+5
y (x) + y (x) + y(x) = 0 , (11.175)
x x2
we get
x−5 x+5
a2 (x) = 1 ; a1 (x) = ; a0 (x) = . (11.176)
x x
The inside integral in (11.174) needs to be done first.

11.10 Solution by Method of Order Reduction 365
% % % %
a1 (x) x−5 5
exp − dx = exp − dx = exp dx − dx
a2 (x) x x

= exp 5 log(x) − x = x5 exp(−x) . (11.177)

Use (11.177) ; insert y1 (x) = σ1 x5 1 − 5x exp(−x) in (11.174); and write the resul-
tant of (11.174) as
%
y2 (x) x x5 exp(−x)
= x 1−
5
exp(−x) !2 dx
(−4σ2 ) 5 x5 exp(−x)(1 − 5x )
%
x exp(x)
= x 1−
5
exp(−x) 2 dx . (11.178)
5 x5 1 − 5x
where (−4σ2 ) is a new arbitrary constant. Power expand the integrand,
exp (x) 1 7 51 389 5

2 = 5 + 4 + 3
+ 2
+ + (x3 ) , (11.179)
x5 1 − 5x x 5x 50x 750x 24x
and integrate (11.179).

%
exp (x) 1 −4 7 51 389 −1
dx = − x − x−3 − x−2 − x
x 2
x5 1− 5 4 15 100 750
5
+ log(x) + O(x2 ) . (11.180)
24

Power expand x5 1 − 5x exp(−x).
x 6 6 7 4
x5 1 − exp(−x) = x5 − x + x7 − x8
5 5 10 15
+ O(x9 ) (11.181)
The right-hand side of (11.178) is equal to the product of the left-hand sides of
(11.181) and (11.180). Below we equate that with the product of the right-hand sides
of (11.180) and (11.181). We get
%
y2 (x) x exp (x)
= x5 1 − exp(−x) 2 dx
(−4σ2 ) 5 x5 1 − 5x
x 1 −4 7 51 389 −1
= x5 1 − exp(−x) − x − x−3 − x−2 − x
5 4 15 100 750
x 5
+ x5 1 − exp(−x) log(x) + O(x2 ) ,
5 24
6 7 1 −4 7 51
= x5 − x6 + x7 − x − x−3 − x−2
5 10 4 15 100
x 5
+ x5 1 − exp(−x) log(x) + O(x2 ) .
5 24
' (
−1 −5 5 x 2x2 x3 2x4
= x 1− exp(−x). log(x) + x + + + + O(x ) .
5
4 6 5 3 2 3
Finally, multiply both sides by −4 to get the second solution of (11.151)-(4) by the
method of order reduction.
y2 (x) y2;4 (x, ν0 = 1)

≡
σ2 σpia;4 ord .reduc.
' (
−5 5 x 2x2 x3 2x4
= x 1− exp(−x). log(x) + x + + + + O(x ) .
5
6 5 3 2 3
(11.182)

Comment: Equation (11.182)—meaning y2;4 (x,ν 0 =1)
σpia;4
—agrees with
ord.reduc.
Piaggio-like second solution, y2;4 (x,ν 0 =1)
σpia;4
, of (11.151)-(4) recorded in (11.171).
piaggio
Complete solution, ycomplete;4 (x), of differential equation (11.151)-(4) is the sum of

the first- and the second solutions given in (11.164) and (11.182).
ycomplete;4 (x)
x
σ1 x 5 1 − exp(−x)
5
' (
−5 5 x 2x2 x3 2x4
+ σ2 x 1− exp(−x). log(x) + x + + + + O(x5 ) .
6 5 3 2 3
(11.183)
11.11 Bessel’s Equation of Order Zero
Indicial Equation Roots Are Equal

Consider (11.184).
11.11 Bessel’s Equation of Order Zero 367
y (x) + y (x) + y(x) = 0 . (11.184)
x x2
Bessel’s equation of order zero is obtained by setting
β = 0, γ = 1, μ = 1, ν = 0 and ρ = 0 in (11.184).
1
y (x) + y (x) + y(x) = 0 . (11.185)
x
It has a regular singular point at x = 0, and its indicial equation has a double root
ν0 = 0. Accordingly, it is a member of category 2. As such the following rule applies.
If the two roots—ν1 and ν2 —of the indicial equation are equal—meaning if
2
ν1 = ν2 ≡ ν0 , or equivalently ρ = γ−1 2
, and the relevant differential equation—
meaning (11.185)—has a regular singular point at x = x0 , then it will always posses a
relatively easily accessible first solution, y1 (x), in the form of modified Taylor series.
∞

y1 (x) = an (x − x0 )(n+ν0 ) . (11.186)
n=0
The second solution, y2 (x), in contrast, may require more effort.

Comment Regarding Procedure for Working out First Solution
Follow the usual procedure for working out the first solution for (11.185) which is
Bessel’s equation of order zero. To that end note, the variables β = 0, γ = 1, μ = 1,
ν = 0, ρ = 0 and the fact that the relevant indicial equation has a double root ν0 = 0.
Next, use (11.85) that relates a1 to a0 . That is
β ν0 + ν 0+0
a1 = −a0 = −a0 = 0 . (11.187)
2 ν0 + γ 0+1
To calculate a2 and higher-order an , for convenience rewrite (11.86) which must

obtain for all n ≥ 2. Even though at the end of these calculations, ν0 has to be set
equal to the double root of the indicial equation—which is ν0 = 0 —for the moment
keep the variable ν0 intact.
an−1 [β (n − 1 + ν0 ) + ν] + an−2 μ
an = −
(n + ν0 )(n + ν0 + γ − 1) + ρ
an−2
= − . (11.188)
(n + ν0 )2
For n = 2, (11.188) gives
a0
a2 = − . (11.189)
(2 + ν0 )2
For n = 3, (11.188) and (11.187) lead to
a3 = 0 (11.190)
For n = 4, (11.188) and (11.189) lead to
a2 a0
a4 = − = (11.191)
(4 + ν0 )2 (4 + ν0 )2 (2 + ν0 )2
and so on. By following the same procedure, one finds for n = 5, 6, 7, 8, 9,
a5 = a7 = a9 ... = 0 , (11.192)
a4 a0
a6 = − = − ,
(6 + ν0 )2 (6 + ν0 )2 (4
+ ν0 )2 (2 + ν0 )2
a6 a0
a8 = − = , (11.193)
(8 + ν0 )2 (8 + ν0 ) (6 + ν0 )2 (4 + ν0 )2 (2 + ν0 )2
2
and so on.
Listed below, in (11.221), is a compact form of the general solution, G BESS0 (x, ν0 ),
of Bessel’s equation, (11.185), of order zero. This general solution is made up of
(11.187) and (11.189)–(11.193), etc.
G BESS0 (x, ν0 )
a0 xν0
1 1
= 1− x2 + x4
(2 + ν0 )2 (4 + ν0 )2 (2 + ν0 )2
1
− x6
(6 + ν0 )2 (4 + ν0 )2 (2 + ν0 )2
1
+ x8
(8 + ν0 )2 (6 + ν0 )2 (4 + ν0 )2 (2 + ν0 )2
1
− x10
(10 + ν0 )2 (8 + ν0 )2 (6 + ν0 )2 (4 + ν0 )2 (2 + ν0 )2
+ O(x12 ) (11.194)
11.11.2 First Solution
The indicial equation for Bessel’s equation of order zero has a double root ν0 =
0. Setting everywhere, in its general solution (11.221), ν0 = 0 and the arbitrary
constant a0 equal to unity leads to the first solution of Bessel’s equation of order
zero. Traditionally, it is labeled as J0 (x).
G BESS0(x,ν0 =0)
= J0 (x)
a0
x2 x4 x6 x8 x10
= 1− + − + − + O(x12 )
(2)2 (8)2 (48)2 (384)2 (3, 840)2
1 x 2 1 x 4 1 x 6 1 x 8
= 1 − 1 + 2 − 2 + 2
1! 2 2! 2 3! 2 4! 2
1 x 10
− 2 + O(x12 ) (11.195)
5! 2
These results are suggestive of the well-known expression
(−1)n x
∞
2n
J0 (x) = . (11.196)
n=0
(n!)2 2
11.11.3 Piaggio-Like Second Solution
As expressed in (11.144), the second solution—that is YBESS0;2;piaggio

(x,
ν0 )—is the differential, with respect to ν0 , of the general solution—that means
d G BESS0 (x,ν0 )
d ν0
. Accordingly,
d G BESS0 (x, ν0 )
YBESS0;2;piaggio (x, ν0 ) ≡
d ν0
2 4(3 + ν0 )
= G BESS0 (x, ν0 ) · log(x) + σ1 xν0 x2 − 2 x4
(2 + ν0 )3 (ν0 + 6ν0 + 8)3

ν0 88 + 48ν0 + 6ν0 2
+ σ1 x x6
(ν03 + 12ν02 + 44ν0 + 48)3
(100 + 70ν0 + 15ν0 2 + ν0 3 )

− σ 1 x ν0 8 x8
(8 + ν0 )3 (6 + ν0 )3 (4 + ν0 )3 (2 + ν0 )3
2(4, 384 + 3, 600ν0 + 1, 020ν0 2 + 120ν0 3 + 5ν0 4 ) 10
− σ 1 x ν0 x
(10 + ν0 )3 (8 + ν0 )3 (6 + ν0 )3 (4 + ν0 )3 (2 + ν0 )3
+ σ1 xν0 O(x12 )... (11.197)
(i) : Replace the arbitrary constants a0 and σ1 in (11.197) by unity everywhere.
(ii) : Knowing that the double root of the relevant indicial equation is ν0 = 0,
replace the symbol ν0 with its value 0 everywhere in (11.197).
Once these changes have been made the second solution, YBESS0;2;piaggio (x, ν0 = 0),
of Bessel’s equation of order zero can be written as
YBESS0;2;piaggio (x, ν0 = 0)
x2 3 11 25
= J0 (x) log(x) + − x4 + x6 − x8
4 128 13, 824 1, 769, 472
137
+ x10 + O(x12 ) . (11.198)
884, 736, 000
Various terms in (11.198) can also be expressed in a form familiar in the literature.
For instance, one can write
x2 1
= (−1)2 2 x2
4 2 (1!)2
3 1 1 4
− x4 = (−1)3 4 1+ x
128 2 (2!)2 2
11 1 1 1 6
x6 = (−1)4 6 1+ + x
13, 824 2 (3!)2 2 3
25 1 1 1 1 8
− x8 = (−1)5 8 1+ + + x
1, 769, 472 2 (4!)2 2 3 4
137 1 1 1 1 1 10
x10 = (−1)6 10 1+ + + + x (11.199)
884, 736, 000 2 (5!)2 2 3 4 5
The results of the Piaggio-like second solution of Bessel’s equation of order zero,
given in (11.199), agree with those in the more elegant (11.200) below.
YBESS0;2;piaggio (x, ν0 = 0)
(−1)n+1 x
∞
2n 1 1 1
= J0 (x) log(x) + 1+ + + ··· + .
n=1
(n!)2 2 2 3 n
(11.200)
As is well known, any linear combination of the first solution—such as (11.196)—

and the second solution—such as (11.200)—of a linear ordinary differential equation
is also a solution of the same differential equation. Accordingly, the following linear
combination of the two is also a second solution. It is traditionally dubbed as Y0 (x),
as the second solution of Bessel’s equation of order zero.
Y0 (x)

(−1)n+1 x
∞
2 2n 1 1 1
= J0 (x) log(x) + 1 + + + ··· +
π n=1
(n!)2 2 2 3 n
2
+ {[γe − log(2)]J0 (x)} . (11.201)
π
Here γe ≈ 0.577215665 is the so-called Euler’s constant.
11.11.4 Solution by Method of Order Reduction
Given the availability of a nonzero solution, y1 (x), of a second-order homogeneous

linear ordinary differential equation a linearly independent second solution can be
determined by a procedure known as order reduction. [Compare (7.31) to (7.45).]
This procedure consists in representing the second solution, y2 (x), as a product of
the first solution, which currently is y1 (x) = J0 (x), and an unknown new function
f (x).
y2 (x) = f (x) y1 (x) . (11.202)
For a differential equation

a2 (x) D2 + a1 (x) D + a0 (x) u(x) = 0 (11.203)
the second solution, y2 (x), is given by the relationship

&
% exp − a1 (x) dx
a2 (x)
y2 (x) = y1 (x) f (x) = σ0 y1 (x) · dx . (11.204)
[y1 (x)]2
Note, σ0 is an arbitrary constant. By comparing (11.203) and the current differential

equation (11.185), written below as (11.205),
1
y (x) + y (x) + y(x) = 0 , (11.205)
x
we get
1
a2 (x) = 1 ; a1 (x) = ; a0 (x) = 1 . (11.206)
x
The inside integral in (11.204) needs to be done first.

% %
a1 (x) 1 1
exp − dx = exp − dx = exp − log(x) = .
a2 (x) x x
(11.207)
Use (11.207) ; insert y1 (x) = J0 (x) in (11.204); and write the resultant as
%
dx
y2 (x) = σ0 J0 (x) . (11.208)
x {J0 (x)}2
Power expand the integrand in (11.208) while using the expansion for J0 (x) given in
(11.196),
1 1 1 5 23 677
= + x+ x3 + x5 + x7
x {J0 (x)}2 x 2 32 576 73, 728
7, 313 218, 491
+ x9 + x11 + O(x13 ) , (11.209)
3, 686, 400 530, 841, 600
and integrate (11.209).

%
dx x2 5 23 677
= log(x) + + x4 + x6 + x8
x {J0 (x)}2 4 128 3, 456 589, 824
7, 317 218, 491
+ x10 + x12 + O(x14 ) .
36, 864, 000 6, 370, 099, 200
(11.210)
Therefore,
%
y2 (x) dx
= J0 (x)
σ0 x {J0 (x)}2
' (
x2 x4 x6 x8
= J0 (x) log(x) + 1 − + − + ×
4 64 2304 147, 456
' 2 (
x 5 23 677 7, 317
+ x +
4
x +
6
x +
8
x 10
4 128 3, 456 589, 824 36, 864, 000
= J0 (x) log(x)
x2 3 11 25
+ − x4 + x6 − x8
4 128 13, 824 1, 769, 472
137
+ x10 . (11.211)
884, 736, 000
Because (11.211) agrees with Piaggio-like second solution of Bessel’s equation of

order zero—recorded in (11.198)—their final solution is the same. [See Ross17. .
Note that Ross’ results for Y0 (x) are identical to those derived by using Piaggio-like
procedure—refer to (11.201).] That is
YBESS0;2;ord .reduc (x, ν0 = 0)

(−1)n+1 x
∞
2n 1 1 1
= J0 (x) log(x) + 1+ + + ··· + . (11.212)
n=1
(n!)2 2 2 3 n
First solution, J0 (x)—given in (11.196)—and the second solution, Y0 (x)—given

in (11.201)—are added together to make complete solution, YBess0comp , of Bessel’s
equation of order zero.
YBess0comp = σ5 J0 (x) + σ6 Y0 (x) (11.213)
where σ5 and σ6 are arbitrary constants.
11.12 Bessel’s Equation of Order nb
Again consider the differential equation
y (x) + y (x) + y(x) = 0 . (11.214)
x x2
A subset of (11.214) is Bessel’s equation of order nb obtained by setting β = 0, γ =

1, μ = 1, ν = 0, and ρ = −n2b .
1 x2 − n2b
y (x) + y (x) + y(x) = 0 . (11.215)
x x2
Clearly, x = 0 is the only singular point for (11.214) and (11.215). And it is a regular
singular point. Recall that a regular singular point at x = 0 would require x × 1x and
x2 × x12 both to be analytic at x = 0: which is clearly the case here. All the other
points are ordinary points.
11.12.1 Indicial Equation
The indicial equation for Bessel’s Equation (11.215) of order nb is
ν02 + ν0 (γ − 1) + ρ = 0 = ν02 + ν0 (1 − 1) − n2b . (11.216)
Its roots are:
ν0 ≡ ν1 = nb ; ν0 ≡ ν2 = − nb . (11.217)
Because one root, −nb , is the negative of the other, nb , either of the two—say nb —can
be assumed to be positive. Depending on the size of nb , Bessel’s equation of order nb
may belong to any one of the three categories: (1), (2), or (3). Description of what these
categories refer to is available contiguous to (11.80)–(11.81), (11.82), and (11.83).
For Bessel’s equation of order nb , β = 0, γ = 1, μ = 1, ν = 0, and ρ = −n2b . The

relevant differential equations are (11.214) and (11.215). For the general solution,
one uses ν0 as a variable. Its numerical value is specified at the end of the process.
Until then proceed as follows.
Use (11.85) that relates a1 to a0 . That is
β ν0 + ν 0+0
a1 = −a0 = −a0 = 0 . (11.218)
2 ν0 + γ 2ν0 + 1
Just like a1 for odd values of n, an = 0.

To calculate a2 and higher-order an , it is helpful to rewrite (11.86) that obtains for
all even values of n that are ≥ 2.
' (
an−1 [β (n − 1 + ν0 ) + ν] + an−2 μ
an = −
(n + ν0 )(n + ν0 + γ − 1) + ρ

an−2
= − . (11.219)
(n + ν0 )2 − n2b
11.12 Bessel’s Equation of Order nb 375
For even values of n, (11.219) provides the recurrence relationship. And for n =
2, 4, 6, 8, etc., it gives
a2 !−1
=− (2 + ν0 )2 − n2b ,
a0
a4 !−1
= (2 + ν0 )2 − n2b (4 + ν0 )2 − n2b
a0
a6 !−1
=− (2 + ν0 )2 − n2b (4 + ν0 )2 − n2b (6 + ν0 )2 − n2b
a0
a8 !−1
= (2 + ν0 )2 − n2b (4 + ν0 )2 − n2b (6 + ν0 )2 − n2b (8 + ν0 )2 − n2b
a0
a10
= −etc. (11.220)
a0
General solution, G BESSnb (x, ν0 ), of Bessel’s equation of order nb , is made up of

(11.218) and (11.220), etc.
G BESSnb (x, ν0 )
a0 xν0

1 1
= 1− x +
2
x4
(2 + ν0 )2 − n2b (2 + ν0 )2 − n2b (4 + ν0 )2 − n2b

1
− x6
(2 + ν0 )2 − n2b (4 + ν0 )2 − n2b (6 + ν0 )2 − n2b

1
+ x8
(2 + ν0 )2 − n2b (4 + ν0 )2 − n2b (6 + ν0 )2 − n2b (8 + ν0 )2 − n2b
+ O(x10 ) (11.221)
11.12.3 Two Solutions
Consider a Bessel’s equation of order nb that belongs to category (1): meaning the
difference, 2 nb , between the two roots—both of which are real—of its indicial
equation is neither zero nor is it equal to an integer. As such, it has two linearly
independent solutions, Bnb (x, ν0 ) which are readily found by setting ν0 = ± nb .
11.12.4 First Solution
For the case when the indicial equation roots difference 2nb is not an integer, the
first solution, Bnb (x, ν0 = nb ), of Bessel’s equation of order nb , is readily found, by
setting ν0 = nb in the general solution G BESSnb (x, ν0 ) given in (11.221).
G BESSnb (x, ν0 = nb ) Bnb (x, ν0 = nb ) x2

= = 1 −
a0 xnb a0 xnb 22 (1 + nb )
x4 x6
+ 4 −
2 .2!(2 + nb )(1 + nb ) 26 .3!(3 + nb )(2 + nb )(1 + nb )
x8
+ 8
2 .4!(4 + nb )(3 + nb )(2 + nb )(1 + nb )
+ O(x10 ) (11.222)
Equation (11.222) is suggestive of the more general expression

∞
Bnb (x, ν0 = nb ) x2 n
= (−1)n
.
a0 x n b 22 n n!(n + nb )(n − 1 + nb )...(3 + nb )(2 + nb )(1 + nb )
n=0
(11.223)
It is tempting to express the denominator of (11.223) by setting
22 n n!(n + nb )(n − 1 + nb )...(3 + nb )(2 + nb )(1 + nb ) .

22 n n!(n + nb )!
≡ (11.224)
nb !
But, alas!, nb is not an integer and its factorial is not defined. Therefore, as currently
written, the relationship (11.224) would be incorrect.
We need to generalize the factorial function to non-integral values of the argument.
This can be done through the use of the gamma function which is defined whenever
the following integrals converge.
% ∞ % 1
(z) = t z−1
exp(−t) dt = (− log t)z−1 dt . (11.225)
0 0
Indeed, (z) is valid for all complex values of z: Be they non-integral, negative or
imaginary.
The simplest property of the gamma function is the recurrence relationship.
(z + 1) = z (z). (11.226)
And for integral z
(z + 1) = z! (11.227)
11.12 Bessel’s Equation of Order nb 377
As such, (11.223) may be expressed as

∞
(nb + 1) x 2n+nb
Bnb (x, ν0 = nb ) = a0 2nb (−1)n . (11.228)
n=0
n! (n + nb + 1) 2
Part of the right-hand side of (11.228) is function Jnb (x) known as the ‘Bessel function
of the first kind of order nb .’
∞
(−1)n x 2n+nb
Jnb (x) = . (11.229)
n=0
n! (n + nb + 1) 2
11.12.5 Second Solution
The second solution Bnb (x, ν0 = −nb ) that refers to ν0 = − nb is now straightforward
to determine. For even values of n, the recurrence relationship is

an−2
an = −
(n − nb )2 − n2b
an−2
=− . (11.230)
n(n − 2 nb )
Hence, the counterparts of (11.222) and (11.223) are
Bnb (x, ν0 = −nb ) x2 x4

= 1 − +
a0 x−nb 22 (1 − nb ) 24 .2!(2 − nb )(1 − nb )
x6
− 6
2 .3!(3 + nb )(2 + nb )(1 + nb )
x8
+ 8
2 .4!(4 − nb )(3 − nb )(2 − nb )(1 − nb )
+ O(x10 ) (11.231)
Equation (11.231) is suggestive of the more general expression

∞

Bnb (x, ν0 = −nb ) x2 n
−n = (−1)n 2 n .
a0 x b 2 n!(n − nb )(n − 1 − nb )...(3 − nb )(2 − nb )(1 − nb )
n=0
(11.232)
It is important to note that because nb in (11.232) is never a whole number, the

denominator in (11.232) does not go to zero. As such there are no pesky divergences
in this equation.
Using the same argument that led from (11.222) to (11.228), we can express
(11.232) by the following.
(−nb + 1) x
∞
2n−nb
Bnb (x, ν0 = −nb ) = a0 2−nb (−1)n (.11.233)
n=0
n! (n − nb + 1) 2
Part of the right-hand side of (11.228) is function Jnb (x) known as the ‘Bessel function
of the first kind of order −nb .’
∞
(−1)n x 2n−nb
J−nb (x) = . (11.234)
n=0
n! (n − nb + 1) 2
Complete solution of differential equation (11.215) for the case where the indicial
equation roots differ by non-integer—meaning the Bessel’s equation is of category
(1)—is the sum of the first- and the second solution that were given in (11.228)
and (11.233).
x 2n+nb
∞
(nb + 1)
Bnb (x, ν0 = nb ) + Bnb (x, ν0 = −nb ) = σ1 (−1)n 2 nb
2 n! (n + nb + 1)
n=0
∞
(−nb + 1) x 2n−nb
+ σ2 2−nb (−1)n . (11.235)
n! (n − nb + 1) 2
n=0
11.13 Bessel’s Indicial Equation
Roots are Equal

If the two roots, nb and −nb , of the indicial equation are equal, they are both zero.
As such the Bessel function is of order zero and has already been analyzed in detail
in (11.185)–(11.213).
11.13 Bessel’s Indicial Equation 379
Roots Differ by Integer

Given that nb is real and positive and Bessel’s indicial equation roots, nb and −nb ,
differ by an integer, there are two possibilities. Either nb is an integer or it is an
half-odd integer.
Roots Are Integers
Differential equations whose indicial equation roots are integers have an important
feature. Their first solution is readily worked out by following the usual technique.
In contrast, their second solution tends to have recurring zeros in the denominator
leading to divergences. Simplest example of this happenstance is a Bessel’s equation
of order unity.
11.13.1 Bessel’s Equation of Order Unity
First Solution
Replacing n2b by unity transforms Bessel’s equation of order nb —see (11.215)—into
Bessel’s equation of order unity—see (11.236).
1 x2 − 1
y (x) + y (x) + y(x) = 0 . (11.236)
x x2
The two roots of the indicial equation of (11.236) are nb = 1 and nb = −1. Its
first solution must use the larger root, namely nb = 1. Because the general solu-
tion G BESSnb (x, ν0 )—see (11.221)—is still applicable, as are other equations leading
to (11.228), all that is needed to convert (11.228) into the first solution, B1;x (nb = 1),
of (11.236) is change the variable nb to unity. This leads to the result
∞
(1 + 1) x 2n+1
B1;x (nb = 1) = a0 21 (−1)n
n=0
n! (n + 1 + 1) 2
∞
(−1) n x 2n+1
= 2 a0 ≡ 2 a0 J1 (x) , (11.237)
n=0
n! (n + 1)! 2
where J1 (x) is Bessel’s function of the first kind of order unity. [Compare (11.229).]
Second Solution
Bessel’s equation of order unity is treated in Piaggio10. , pp. 114–115. Because the
Piaggio book is not readily available, and the presentation there is minimal and
somewhat abstruse, it is helpful to record a detailed solution here.
Consider the general solution, (11.221), of Bessel’s equation of order nb and

convert it to general solution of Bessel’s equation of order unity by setting nb = 1.
G BESS1 (x, ν0 )
' ( ' (
x2 x4
= a0 xν0 [1 − +
(ν0 + 3)(ν0 + 1) (ν0 + 5)(ν0 + 3)2 (ν0 + 1)
' (
x6
−
(ν0 + 7)(ν0 + 5)2 (ν0 + 3)2 (ν0 + 1)
' (
x8
+
(ν0 + 9)(ν0 + 7)2 (ν0 + 5)2 (ν0 + 3)2 (ν0 + 1)
+ O(x10 )] (11.238)
When ν0 = 1, (11.238) is the first solution, B1;x (nb = 1), of Bessel’s equation of order
unity. [Reminder: The first solution, B1;x (nb = 1), of Bessel’s equation of order unity
is (11.237).] On the other hand, when ν0 is set equal to −1, the denominator of the
general solution (11.238) goes to zero because of the presence of the factor (ν0 + 1)
everywhere. To avoid this difficulty, Piaggio proposed to change the arbitrary constant
a0 to another arbitrary constant a0 (ν0 + 1). The Piaggio10. version of the general
solution is
G BESSpiaggio (x, ν0 )
' (
ν0 x2
= a0 x [(ν0 + 1) −
(ν0 + 3)
' 4 (
x
+
(ν0 + 5)(ν0 + 3)2
' (
x6
−
(ν0 + 7)(ν0 + 5)2 (ν0 + 3)2
' (
x8
+
(ν0 + 9)(ν0 + 7)2 (ν0 + 5)2 (ν0 + 3)2
+ O(x10 )] (11.239)
The important part of Piaggio’s suggestion is the following: the second solution,
B2;x (nb ), of Bessel’s equation of order unity is the differential with respect to the
variable ν0 of its general solution G BESSpiaggio (x, ν0 ). The final requirement is that
the result of such differential be evaluated at ν0 = −1 which is the second root of
the relevant indicial equation. Recall (11.121), whereby
dxν0
= xν0 log(x) , (11.240)
d ν0
11.13 Bessel’s Indicial Equation 381
and proceed as follows:
d G BESSpiaggio (x, ν0 )
= log(x) G BESSpiaggio (x, ν0 )
d ν0
' ( ' (
1 13 + 3ν0
+ a0 xν0 [1 + x 2
− x4
(ν0 + 3)2 (ν0 + 5)2 (ν0 + 3)3
' (
127 + 52ν0 + 5ν02
− x6
(ν0 + 7)2 (ν0 + 5)3 (ν0 + 3)3
' (
1383 + 753ν0 + 129ν02 + 7ν03
− x8
(ν0 + 9)2 (ν0 + 7)3 (ν0 + 5)3 (ν0 + 3)3
+ O(x10 ) (11.241)
Setting ν0 = −1 leads to the second solution B2;x (nb = −1).
B2;x (nb = −1) = log(x) B1;x (nb = 1)

a x2 5 5 47
0
+ 1+ − x4 + x6 − x8 + O(x10 ).
x 4 64 1152 442, 368
(11.242)
Chapter 12
Answer to Assigned Problems
12.1 Problems Group I, 3-chapt
[Note: Look just below (3.42).]

Scomp for problems (1)–(10)
σ0 exp(x) + σ1 exp(−3x) (1)

σ0 exp(4x) + σ1 exp(−x) (2)
x
(σ0 + σ1 x) exp − (3)
2x
(σ0 + σ1 x) exp (4)
2
(σ0 + σ1 x + σ2 x 2 ) exp (−5x) (5)
(σ0 + σ1 x + σ2 x 2 ) exp (2x) (6)
√ √
x 3 3
exp − σ1 sin x + σ2 cos x (7)
2 2 2
√ √
x 3 3
exp σ1 sin x + σ2 cos x (8)
2 2 2
√ √
exp (−x) σ1 sin 2 x + σ2 cos 2x (9)
√ √
3 7 7
exp − x σ1 sin x + σ2 cos x (10)
2 2 2
12.2 Problems Group II, 3-chapt
[Note: Look above (3.74).]

https://doi.org/10.1007/978-3-319-76406-1_12
384 12 Answer to Assigned Problems
I pi for problems (1)–(10)
c0 c1
(1) − − 40 + 42x + 18x 2 + 9x 3
3 27
c0 c2
(2) − + 153 − 156x + 72x 2 − 32x 3
4 128
(3) 4c0 + 4c3 −192 + 72x − 12x 2 + x 3
(4) 4c0 + 4c4 1920 + 768x + 144x 2 + 16x 3 + x 4
c0 c5
(5) + 45 − 60x + 36x 2 − 12x 3 + 2x 4
8 16
c0 c6
(6) − − 45 + 60x + 36x 2 + 12x 3 + 2x 4
8 16
(7) c0 + c7 −120x + 60x 2 − 5x 4 + x 5
(8) c0 + c8 −120x − 60x 2 + 5x 4 + x 5
81c0 c9
(9) + 400 − 1320x + 720x 2 + 180x 3 − 270x 4 + 81x 5
243 243
256c0 c10
(10) + −4095 = 8100x − 3960x 2 − 1440x 3
1024 1024
c10
+ 2400x 4 − 1152x 5 + 256x 6
1024
12.3 Problems Group III, 3-chapt

Particular integrals, I pi , for problems (1)–(10) are given below.
c0 c1 exp(αt)
− + 2 . (1)
3 α + 2α − 3
c0 c2 exp(αt)
− + 2 . (2)
4 α − 3α − 4
4c3 exp(αt)
4c0 + 2 . (3)
4α + 4α + 1
4c4 exp(αt)
4c0 + 2 . (4)
4α − 4α + 1
c0 c5 exp(αt)
+ 3 . (5)
8 α + 6α2 + 12α + 8
c0 c6 exp(αt)
− + 3 . (6)
8 α − 6α2 + 12α − 8
12.3 Problems Group III, 3-chapt 385
c7 exp(αt)
c0 + . (7)
α2 + α + 1
c8 exp(αt)
c0 + 2 . (8)
α −α+1
c0 c9 exp(αt)
+ 2 . (9)
3 α + 2α + 3
c0 c10 exp(αt)
+ 2 . (10)
4 α + 3α + 4
12.4 Problems Group IV, 3-chapt

sin(x) − 2 cos(x)
. (1)
5
−i
[5 sin(x) − 3 cos(x)] . (2)
17
8 24 6 70
sin(x) − sin(3x) − cos(x) + cos(3x) . (3)
25 1369 25 1369
6 70 8 24
sin(x) − sin(3x) − cos(x) + cos(3x) . (4)
25 1369 25 1369
1 sin(2 x) − cos(2 x)
+ . (5)
4 16
1 sin(2 x) + cos(2 x)
+ . (6)
4 16
4 sin(2x) − 6 cos(2x)
. (7)
13
2i
− [3 sin(2x) − 2 cos(2x)] . (8)
13
2 16 sin(4x) − 26 cos(4x)
+ . (9)
3 233
1 sin(4x) − cos(4x)
− + . (10)
2 12
12.5 Problems Group V, 3-chapt

exp(2x)
{[−4i + 39x] sin(x) + [3 + 26x] cos(x)} . (1)
169
−i exp(2x)
{[2 + 35x] sin(x) + [11 + 5x] cos(x)} . (2)
125
2 2
[3 sin(x) + 4 cos(x)] − [35 sin(3x) + 12 cos(3x)] . (3)
25 1369
2 2
− [4 sin(x) + 3 cos(x)] + [12 sin(3x) + 35 cos(3x)] . (4)
25 1369
exp(x
− {72 sin(x) + 21 cos(x) − x[65 sin(x) + 45 cos(x)]} . (5)
625
exp(x){3 sin(x) + x [sin(x) − cos(x)]} . (6)
2 2 exp(x)
exp(x)(1 − x) − [(104 − 222x) sin(2x) − (153 − 37x) cos(2x)] . (7)
3 1369
2 exp(x)
−2 exp(x)(1 − x) + [(32 + 26x) sin(2x) + (43 − 39x) cos(2x)] . (8)
169
4 exp(−x)[2 sin(x) + x cos(x)] . (9)
2 exp(x)
[(130 − 259x) sin(x) − (151 − 185x) cos(x)] . (10)
1369
12.6 Problems Group VI, 3-chapt
[Note: Look at (3.143).]

Given below are the separated simultaneous linear ordinary equations, their Scomp
and I pi for problems (1)–(3), and some help with the solution of problems (4)–(6).
(22 + 2 + 1)x = 6 exp(t) − 3 t exp(t) ,

(22 + 2 + 1)y = 2 exp(t) t − 2 exp(t) .

t t t
Scomp; x = exp − σ1 sin + σ2 cos ,
2 2 2

t t t
Scomp; y = exp − σ3 sin + σ4 cos .
2 2 2

3 16 2 11
I pi ; x = exp(t) − t , I pi ; y = exp(t) − + t ;
5 5 5 5
1 1
σ3 = − (2σ1 − σ2 ) ; σ4 = − (σ1 + 2σ2 ) . (1)
5 5
12.6 Problems Group VI, 3-chapt 387
(32 − 3)x = 2 exp(t) + 3 t exp(t) − 2 t − 1 ,

(3 − 3)y = − exp(t) − 3 t exp(t) + t + 2 .
2
Scomp; x = σ1 exp (−t) + σ2 exp (t) ,

Scomp; y = σ3 exp (−t) + σ4 exp (t) ,
exp(t) 1
I pi ; x = −1 + 2 t + 6 t 2 + (1 + 2 t) ,
24 3
exp(t) 1
I pi ; y = −1 + 2 t − 6 t 2 − (2 + t) ;
24 3
σ3 = σ1 ; σ4 = − σ2 . (2)
( + 2 + 2)x = 1 + 2 t − exp(t) − 2 t exp(t) ,
2
(2 + 2 + 2)y = 3 exp(t) + 3 t exp(t) − 1 .

Scomp; x = exp (−t) [σ1 sin(t) + σ2 cos(t)] ,
Scomp; y = exp (−t) [σ3 sin(t) + σ4 cos(t)] ,
exp(t) 1
I pi ; x = (3 − 10 t) − + t ,
25 2
exp(t) 1
I pi ; y = (3 + 15 t) − ;
25 2
σ3 = σ1 ; σ4 = σ2 . (3)
In problem (4) set (−x + 3) = exp(−t) , (−x + 3)D = , and thereby trans-
form it into problem (1).
In problem (5) set (2x − 1) = exp(2t) , (2x − 1)D = , and thereby trans-
form it into problem (2).
In problem (6) set (x + 1) = exp(t) , (x + 1)D = and (x + 1)2 D 2 =
( − 1) and thereby transform it into problem (3).

Solution u = u(x) to problem set I problems (1)–(12) is given below.
3
1 x
u(x) = + σ0 . (1)
x 3
6
1 x
u(x) = 3 + σ0 . (2)
x 6
1
u(x) = 3 exp(x)(x − 1) + σ0 . (3)
x2 exp(x)

1 exp(4x)
u(x) = (8x 2 − 4x + 1) + σ0 . (4)
x exp(3x) 32
1
u(x) = − log(cos x) + σ0 . (5)
sin x
u(x) = (cos x) [x + σ0 ] . (6)

1 1
u(x) = exp(2x)(x − 1)2 + σ0 . (7)
(x − 1) exp(x) 2

1 (sin x)4
u(x) = + σ0 . (8)
(sin x)3 2
2
(x + 2x)
u(x) = (x + 1) + σ0 . (9)
2
2
(x − 2x)
u(x) = (x − 1) + σ0 . (10)
2
2
1 2x 2 (log x)2 − 2x 2 log(x) + x 2
u(x) = + σ0 . (11)
log x 4
1
u(x) = [x sin(x) + cos(x) + σ0 ] . (12)
x(sin x)

Solution u(x) to problem set II, problems (1)–(10) is given below.
21
1
u(x) = ± (2 + σ0 exp (−6 x)) . (1)
3
2
u(x) = . (2)
3 + σ0 exp(x 2 )
21
3
u(x) = ± . (3)
2 + σ0 exp (3 x 2 )
21
1
u(x) = ± . (4)
exp(x 2 ){−6 x + σ0 }
13
1
u(x) = ; (5)
exp(x 3 ){−18 x + σ0 }
13
2 1
u(x) = (−1) 3 ; (5)
exp(x 3 ){−18 x + σ0 }
12.8 Problems Group II, 4-chapt 389
13
4 1
u(x) = (−1) 3 . (5)
exp(x ){−18 x + σ0 }
3
1
u(x) = ± 2 x 2 − 1 + σ0 exp(−2x 2 ) 4
; (6)
1
u(x) = ± (i) 2 x − 1 + σ0 exp(−2x )
2 2 4
. (6)
1
u(x) = . (7)
x(−3 x + σ0 )
1
u(x) = 2 . (8)
x [−2 log(x) + σ0 ]
21
1
u(x) = ± . (9)
2x 2 + σ0 x 4

3
9 x 8 + σ1
u(x) = ; (10)
2 x2
2
3
9 x 8 + σ1
u(x) = (−1) 3 ; (10)
2 x2
4
3
9 x 8 + σ1
u(x) = (−1) 3 . (10)
2 x2

General Solution
y = x σ + 3 σ 2 . (1)
y = x σ + 3 σ 3 . (2)
y = x σ − 2 sin σ . (3)
y = x σ + 2 cos σ . (4)
Singular Solution
12 y + x 2 = 0 . (1)
81 y 2 + 4 x 3 = 0 . (2)
x 2
4 − x 2 = y − x cos−1 . (3)
2
x 2
4 − x 2 = y − x sin−1 . (4)
2

General Solution
√
x = σ1 q − 18 q . (1)
3
q2
y = σ1 − 9 q2 . (1)
3
1
q
x = σ1 q 4 − 2 . (2)
5 3
q4 q2
y = σ1 − . (2)
5 3

σ1 5
x = − . (3)
q3 q
σ1
y = 3 √ + 5 log(q) − 3 . (3)
q
Singular Solution
Singular solution to (1), (2), and (3) is the same. That is:
y(x) = 0 .

dy dx
(1) : = ; log(y) = log(x) + const .
y x
Or Equivalently : y = σ0 x .

dy exp(−3y) exp(−2x)
(2) : = exp(−2x) dx ; = + const .
exp(3y) −3 −2

1 3
Or Equivalently : y = − log exp(−2x) + σ0 .
3 2

dy
(3) : = 2x exp(x 2 ) dx ; log(y) = exp(x 2 ) + σ0 .
y
Or Equivalently : y = log−1 [exp(x 2 ) + σ0 ] .

dy
(4) : = x 2 dx ; log(y) = exp(x 2 ) + σ0 .
y2
Or Equivalently : y = log−1 [exp(x 2 ) + σ0 ] .
[Note: Look just below (6.41)]

x
(1) : y = −x + .
1 − σ0 x
(2) : y= x sin−1 (σ0 x)3 .
(3) : y= 5 x σ0 + log x .
√
(4) : y= ± x 8 σ0 + log x .

(5) : 4y = x 3 log(σ0 x) ± {3 log(σ0 x)}2 + 8 .
(1) : y = (x + 1) tan {log(x + 1) + σ0 } .

(2) : (y + 2) = (x + 1)[log(x + 1) + σ0 ] .
(3) : y = (x + σ0 ) cos x .
(4) : (y + 2) = (x + 3) log[σ0 + log(x + 3)] .

(5) : 2 y = − x − 3 ± 3 x 2 + 10 x + σ0 .

(6) : 3 y = − x − 4 ± 4 x 2 + 14 x + σ0 .
1
(7) : 3 y = x + log [2(x + y) + 3] + σ0 .
2
12.14 Problems Group VI, 6-chapt

σ3 − x
(1) : y = .
x2 + 2
(2) : σ3 = x(1 + y + y 2 ) : or equivalently,

1 4σ3
y = −1 ± −3 + .
2 x
x2 y2
(3) : σ3 = + + x exp(y) .
2 2
12.15 Problems Group VII, 6-chapt
[Note: See (6.241).]

Procedure: First choose u(x, y) and v(x, y) from one of the three equations
in problem set V I chapter (ST). Then, just to be sure, check, by using (6.82), as
to whether the given differential equation is exact or inexact. And if it is inexact,
employ either (6.101) or (6.103) to find whether an integrating factor that depends
only upon a single variable is possible. When the equation has been made exact,
solve it according to (6.88) and (6.91).
σ0 2
(1) : y = 3 − .
(2x + 3) 2 3
σ0
(2) : y = √ .
x
3x 2 2
(3) : y = σ0 exp − + − x2 .
2 3
12.16 Problems Group VIII, 6-chapt
1
(1) : y = −1 +
x + σ0
√ √
1 7 7
(2) : y=− + tan σ0 − x .
2 2 2
12.17 Problems Group IX, 6-chapt

12.17 Problems Group IX, 6-chapt 393
Given below are the Scomp and I pi for problems (1)–(5) and some help with the
solution.
Set (x + 3) = exp(t) ; (x + 3) D = ; (x + 3)2 D 2 = ( − 1) .

T her e f or e, (1) : (2 + )u = − t exp(t) ,

3 t
Scomp; t = σ0 + σ1 exp(−t) , I pi; t = exp(t) − ,
4 2

σ1 3 log(x + 3)
Scomp; x = σ0 + , I pi; x = (x + 3) − , (1)
x +3 4 2
and (2) : (2 + + 3)u = − t exp(t) ,
√ √
t 13 13
Scomp; t = exp − σ1 sin t + σ2 cos t ,
2 2 2

3 t
I pi; t = exp(t) − ,
25 5
√ √
1 13 13
Scomp; x = √ σ1 sin log(x + 3) + σ2 cos log(x + 3)
x +3 2 2

3 log(x + 3)
I pi; x = (x + 3) − . (2)
25 5
Set (2x − 1) = exp(2t) ; (2x − 1) D = ; (2x − 1)2 D 2 = ( − 1) .

T her e f or e, (3) : (2 − 4)u = − 4t exp(2t) sin(2t) ,
Scomp; t = σ0 + σ1 exp(4t) ,
exp(2t)
I pi; t = [cos(2t) + 2t sin(2t)] ,
4
Scomp; x = σ0 + σ1 (2x − 1)2 ,
(2x − 1)
I pi; x = {cos[log(2x − 1)] + log(2x + 1) sin[log(2x − 1)]} , (3)
4
and (4) : (2 − 4 + 4)u = −4t exp(2t) sin(2t) ,
Scomp; t = exp (2t) [σ1 t + σ0 ] ,
I pi; t = exp(2t) [cos(2t) + t sin(2t)] ,
Scomp; x = (2x − 1) σ2 log(2x − 1) + σ4 ,
(2x − 1)
I pi; x = 2 cos(log(2x − 1)) + log(2x − 1) sin(log(2x − 1)) . (4)
2
Set (3x + 2) = exp(3t) ; (3x + 2) D = ; (3x + 2)2 D 2 = ( − 3) .

1
T her e f or e, (5) : − 6 +
2
u = exp(3t) sin(3t) ,
4
√ √
35 35
Scomp; t = exp(3t) σ1 exp t + σ2 exp − t ,
2 2
4 exp(3t)
I pi; t = − sin(3t) ,
√
71 √
+ σ2 (3x + 2)−
35 35
Scomp; x = (3x + 2) σ1 (3x + 2) 6 6 ,
4(3x + 2)
I pi; x = − sin[log(3x + 2)] . (5)
71
x3
(1) : y= − x + σ0 x 2 + σ0 x + σ1
12
x3
or , y = − x + σ0 −x 2 + σ0 x + σ1 .
12
(1) : y = 2 σ0 tanh [σ0 x + σ1 ]

x
(2) : y = − log σ0 − σ1 exp − .
σ0
(1) : u 2 (x) = σ2 cos(x)

√
(2) : u 2 (x) = σ2 exp (− 2 − 1) x
x x
(3) : u 2 (x) = σ2 exp − cos
2 2
(4) : u 2 (x) = σ2 exp (x) cos (x)
√
(5) : u 2 (x) = σ2 x (1− 2)

x2 −x 2
(6) : u 2 (x) = σ2 exp · exp dx
2 2

x2 x 2 + 4x
(7) : u 2 (x) = σ2 exp −x − · exp .
2 2
exp(x)
(1) : I pi (x) =
2
x2 3
(2) : I pi (x) =− +x+
2 2

2 2
(3) : I pi (x) = x−
3 3
(4) : I pi (x) = log(x) + 1
1
(5) : I pi (x) =
2x
x
(6) : I pi (x) = log(x) − 1
2
1
(7) : I pi (x) = 2 log(x) + 1
8x
(8) : I pi (x) = (x + 1) log(x + 1) − (x + 1) log(x) − 1
1
(9) : I pi (x) =
2
x 1
(10) : I pi (x) = + .
2 2

Using the procedure outlined in (7.64)–(7.71), solution of differential equation
obeyed by y is given below: as is the solution u.
√ √ 1
(1) : y(x) = σ1 sin(x 5) + σ2 cos(x 5) + ; u(x) = exp(x 2 ) y(x).
5
x2
(2) : y(x) = σ1 sin(x) + σ2 cos(x) + 1 ; u(x) = exp y(x).
2
x √ x √ 2
(3) : y(x) = σ1 sin 6 + σ2 cos 6 + ;
2 2 3
3x 2
u(x) = exp y(x).
4
√ √
2+ 2 + σ2 (x) 2 − + x −1 ; u(x) = x y(x).
1 5 1 5
(4) : y(x) = σ1 (x) 2

All we have to do to solve (11.31) is to use in (11.24) their given values of α,
β, γ. [Note: These symbols were defined in (11.5).] First few terms of the relevant
Frobenius power series solution of (11.31)-(1)–(10) are the following.
For (1), α = 1, β = −1, γ = 1. Therefore

1 5 37
(1) : u(x) = σ1 1 − x +
2
x −
4
x + O(x )
6 8
2 24 720

1 7 31
+ σ2 x − x +
3
x −
5
x + O(x ) .
7 9
3 60 1260
For (2) , α = 1, β = −3, γ = 2. Therefore

7 13
(2) : u(x) = σ1 1 − x +
2
x −
4
x + O(x )
6 8
12 60

1 11 137
+σ2 x − x +
3
x −
5
x + O(x ) .
7 9
2 40 1680
For (3) , α = 1, β = 3, γ = −2. Therefore

1 1
(3) : u(x) = σ1 1 + x 2 − x4 − x 6 + O(x 8 )
4 12

1 19 1
+ σ2 x + x3 − x5 − x 7 + O(x 9 ) .
6 120 1680
For (4) , α = 4, β = −1, γ = −4. Therefore

7 3
(4) : u(x) = σ1 1 + 2 x −
2
x +
4
x + O(x )
6 8
12 10

1 2
+ σ2 x + x −
5
x + O(x ) .
7 9
20 105
For (5) , α = −1, β = −1, γ = 1. Therefore

12.23 Problems Group I, 11-chapt 397

1 1 1
(5) : u(x) = σ1 1 − x2 + x4 − x 6 + O(x 8 )
2 24 80

1 1
+ σ2 x + x5 + x 7 + O(x 9 ) .
20 210
For (6) , α = 1, β = 1, γ = 1. Therefore

1 5 37
(6) : u(x) = σ1 1 + x +
2
x +
4
x + O(x )
6 8
2 24 720

1 7 1
+ σ2 x + x +
3
x +
5
x + O(x ) .
7 9
3 60 180
For (7) , α = −1, β = −1, γ = −1. Therefore

1 5 37
(7) : u(x) = σ1 1 + x +
2
x +
4
x + O(x )
6 8
2 24 720

1 7 31
+ σ2 x + x +
3
x +
5
x + O(x ) .
7 9
3 60 1260
For (8) , α = 2, β = 2, γ = 2. Therefore

1 2
(8) : u(x) = σ1 1 − x 2 + x4 − x 6 + O(x 8 )
3 45

2 1 1
+ σ2 x − x3 + x5 − x 7 + O(x 9 ) .
3 6 63
For (9) , α = −2, β = −2, γ = −2. Therefore

2 13
(9) : u(x) = σ1 1 + x +
2
x +
4
x + O(x )
6 8
3 45

2 11 43
+ σ2 x + x +
3
x +
5
x + O(x ) .
7 9
3 30 315
For (10) , α = 3, β = 3, γ = −3. Therefore

3 5 3
(10) : u(x) = σ1 1 + x −
2
x +
4
x + O(x )
6 8
2 8 80

3 3
+ σ2 x − x5 + x 7 + O(x 9 ) .
20 70

Information given in (11.52)–(11.58) is used to solve the five differential equations
in problem set (II). These problems are similar in form to (b)-(11.32).
For (1) , α = 1, β = 2, γ = 1, μ = 3, ν = 3, ρ = 1. Therefore

x2 x3 x4 7
(1) : u(x) = σ1 1 − − + + x 5
2 3 24 30

47 135 x2 x3 x4
+ σ1 x −
6
x + σ2 x −
7
− −
720 2520 2 3 24
5
x 81 29
+ σ2 + x6 − x 7 + O(x 8 ) .
10 720 2520
For (2) , α = 0 ; β = 1 ; γ = 0 ; μ = 2 ; ν = 2 ; ρ = 0. Therefore

x3 x4 x5 2 18
(2) : u(x) = σ1 1 − − + + x +
6
x 7
3 6 20 45 1008

x3 x4 3 x6 25
+ σ2 x − − +− x5 + + x 7 + O(x 8 ) .
6 6 40 30 1008

x2 x4 x5 5 x7
(3) : u(x) = σ1 1 − − + + x +
6
2 8 10 240 105
7
x2 x4 9 x
+ σ2 x − − + x6 + + O(x 8 ) .
2 8 240 105

x2 x3 x4 2 13 23
(4) : u(x) = σ1 1 − + + − x5 + x6 + x7
2 3 24 15 504 720
3 4
x x 23 61 13
+ σ2 x − x 2 + + − x5 + x6 − x7
6 3 120 1008 360
+ O(x 8 ) .

x3 11 7 x6 1087
(5) : u(x) = σ1 1 − 2x 2 + + x4 + x5 + − x7
2 24 30 18 5040

x2 2 x4
+ σ2 x − − x3 +
2 3 6
12.24 Problems Group II, 11-chapt 399

3 131 8
+ σ2 x5 + x6 + x 7 + O(x 8 ) .
40 720 5040

We use (11.85), set n = 2, 3, 4, 5, 6, 7 in (11.86), and solve the following differ-
ential equations that are of the form (11.78). Only the indices β, γ, ν, ρ were given
in (11.97)-(1)–(5). For convenience, they are reprinted below.
(1) : [β = −3 ; γ = 3 ; ν = 2 ; ρ = −4 ] .
(2) : [β = 2 ; γ = −3 ; ν = 1 ; ρ = −2 ] .
(3) : [β = 3 ; γ = −2 ; ν = −2 ; ρ = −3 ] .
(4) : [β = 1 ; γ = −2 ; ν = −1 ; ρ = −1 ] .
(5) : [β = −1 ; γ = −5 ; ν = −1 ; ρ = −4 ] .
For (1) , β = −3 ; γ = 3 ; ν = 2 ; ρ = −4 and the solutions are

u(x)
(1a) : √
σ1 x −1+ 5
x √ x2 √ x3 √
= 1+ (35 − 13 5) + (230 − 99 5) + (9, 675 − 4, 283 5)
19 76 2, 508
x4 √ x5 √
+ (110, 555 − 49, 331 5) + (15, 306 − 6, 805 5)
20, 064 25, 080
x6 √ x7 √
+ (23, 965 − 10, 671 5) + (432, 160 − 192, 689 5) + · · ·
109, 440 5, 554, 080

u(x)
(1b) : √
σ1 x −1− 5
x √ x2 √ x3 √
= 1+ (35 + 13 5) + (230 + 99 5) + (9, 675 + 4, 283 5)
19 76 2, 508
x4 √ x5 √
+ (110, 555 + 49, 331 5) + (15, 306 + 6, 805 5)
20, 064 25, 080
x6 √ x7 √
+ (23, 965 + 10, 671 5) + (432, 160 + 192, 689 5) + · · ·
109, 440 5, 554, 080
For (2) , β = 2 ; γ = −3 ; ν = 1 ; ρ = −2 and the solutions are


u(x)
(2a) : √
σ1 x 2+ 6
x √ x2 √ x3 √
= 1− (19 + 8 6) + (67 + 27 6) − (581 + 241 6)
23 92 1, 380
x4 √ x5 √
+ (4, 312 + 1, 657 6) + (−3, 544 + 211 6)
22, 080 22, 080
x6 √ x7 √
+ (42, 782 − 12, 953 6) + (−224, 746 + 83, 447 6) + · · ·
264, 960 1, 854, 720

u(x)
(2b) : √
σ1 x 2− 6
x √ x2 √ x3 √
= 1− (−19 + 8 6) + (67 − 27 6) + (−581 + 241 6)
23 92 1, 380
x4 √ x5 √
+ (4, 312 − 1, 657 6) − (3, 544 + 211 6)
22, 080 22, 080
x6 √ x7 √
+ (42, 782 + 12, 953 6) − (224, 746 + 83, 447 6) + · · ·
264, 960 1, 854, 720
For (3) , β = 3 ; γ = −2 ; ν = −2 ; ρ = −3 and the solutions are

⎡ ⎤
u(x)
(3a) : ⎣ √ ⎦
3+ 21
σ1 x 2
x √ x2 √ x3 √
= 1− (29 + 21) + (647 + 93 21) − (5, 847 + 1, 573 21)
20 680 12, 240
x4 √ x5 √
+ (114 + 11 21) + (−3, 057 + 347 21)
288 5, 760
x6 √ x7 √
+ (95, 543 − 17, 163 21) + (−1, 021, 789 + 205, 549 21) + · · ·
172, 800 2, 419, 200
⎡ ⎤
u(x)
(3b) : ⎣ √ ⎦
3− 21
σ1 x 2
x √ x2 √ x3 √
= 1+ (−29 + 21) + (647 − 93 21) + (−5, 847 + 1, 573 21)
20 680 12, 240
x4 √ x5 √
+ (114 − 11 21) − (3, 057 + 347 21)
288 5, 760
x6 √ x7 √
+ (95, 543 + 17, 163 21) − (1, 021, 789 + 205, 549 21) + · · ·
172, 800 2, 419, 200
For (4) , β = 1 ; γ = −2 ; ν = −1 ; ρ = −1 and the solutions are


u(x)
(4a) : √
3+ 13
σ1 x 2
x x2 √ x3 √
= 1 − + (7 + 13) − (1 + 13)
2 72 144
x4 √ x5 √
+ (−2 + 13) + (29 − 10 13)
288 8, 640
x 6 √ x7 √
+ (−81 + 25 13) + (689 − 203 13) + · · ·
88, 320 3, 709, 440

u(x)
(4b) : √
3− 13
σ2 x 2
x x2 √ x3 √
= 1 − − (−7 + 13) + (−1 + 13)
2 72 144
x4 √ x5 √
− (2 + 13) + (29 + 10 13)
288 8, 640
x6 √ x7 √
− (81 + 25 13) + (689 + 203 13) + · · ·
88, 320 3, 709, 440
For (5) , β = −1 ; γ = −5 ; ν = −1 ; ρ = −4 and the solutions are

u(x)
(5a) : √
σ1 x 3+ 13
x √ x2 √ x3 √
= 1 + (22 + 7 13) + (15 + 4 13) + (588 + 167 13)
51 68 8, 772
x4 √ x5 √
+ (10, 267 + 2, 773 13) + (3, 121 + 915 13)
631, 584 1, 052, 640
x6 √ x7 √
+ (407 + 87 13) + (−3, 205 + 1, 463 13) + · · ·
743, 040 15, 603, 840

u(x)
(5b) : √
σ1 x 3− 13
x √ x2 √ x3 √
= 1 + (22 − 7 13) + (15 − 4 13) + (588 − 167 13)
51 68 8, 772
x4 √ x5 √
+ (10, 267 − 2, 773 13) + (3, 121 − 915 13)
631, 584 1, 052, 640
x6 √ x7 √
+ (407 − 87 13) − (3, 205 + 1, 463 13) + · · ·
743, 040 15, 603, 840
Chapter 13
Answer to Additional Assigned Problems
13.1 Fourier Transforms
A function f (t) that is not necessarily periodic but is reasonably well behaved can
be represented in terms of an integral involving its Fourier transform.
∞
1
f (t) = F(ω) exp(i ω t)dω , (13.1)
2π −∞
where
∞
F(ω) = f (t) exp(−i ω t)dt . (13.2)
−∞
Often F(ω) and f (t) are referred to as the inverse Fourier transform of each other.
13.2 Examples Set (I) and Solution
Work out Fourier transform of six functions given below.
c1 f 1 (t) + c2 f 2 (t) . (1)

f ∗ (t) . (2)
f (t) . (3)
For α > 0 , f (α t) . (4)
f (t − t0 ) . (5)
exp(iω0 t) f (t) . (6) (13.3)

https://doi.org/10.1007/978-3-319-76406-1_13
404 13 Answer to Additional Assigned Problems
13.3 Solution to Examples Set (I) Fourier transforms

∞
(1) : Consider c1 f 1 (t) + c2 f 2 (t) = (1/(2π) −∞ [c1 F1 (ω) + c2 F2 (ω)] exp(i ω t)
dω. The right-hand side of the above equation leads to the conclusion that the rele-
vant Fourier transform for c1 f 1 (t) + c2 f 2 (t) is
c1 F1 (ω) + c2 F2 (ω). (1)
∞
(2) : Consider f (t) = 1/(2π) −∞ F(ω) exp(i ω t) dω.
Take complex conjugate
∞ of both sides of the above equation.
f ∗ (t) = 1/(2π) −∞ F ∗ (ω) exp(−i ω t) dω. Set ω = −ω0 .
−∞
= −1/(2π) ∞ F ∗ (−ω0 ) exp(i ω0 t) dω0 . Now set ω0 = ω. We get
∞
f ∗ (t) = 1/(2π) −∞ F ∗ (−ω) exp(i ω t) dω.
Therefore, the relevant Fourier transform for f ∗ (t) is F ∗ (−ω). (2)
∞
(3) : Consider f (t) = 1/(2π) −∞ F(ω) exp(i ω t) dω.
Differentiate both∞sides with respect to t.
f (t) = 1/(2π) −∞ (iω)F(ω) exp(i ω t) dω.
Therefore, the relevant Fourier transform for f (t) is i ω F(ω). (3)
∞
(4) : Consider f (t ) = 1/(2π) −∞ F(ω ) exp(i ω t ) dω .
∞
Use t = α t and write f (α t) = 1/(2π) −∞ F(ω ) exp(i ω α t) dω .
∞
Setting ω α = ω gives f (α t) = 1/(2π) −∞ F(ω/α) exp(i ω α t) dω/α.
Therefore, the relevant Fourier transform for f (α t) where α > 0 is F(ω/α)/α. (4)
∞
(5) : Consider f (t ) = 1/(2π) −∞ F(ω) exp(i ω t ) dω.
∞
Use t = t − t0 and write f (t − t0 ) = 1/(2π) −∞ F(ω) exp[i ω (t − t0 )] dω.
∞
Rewriting gives: f (t − t0 ) = 1/(2π) −∞ [F(ω) exp(−iω t0 )] exp(i ω t dω.
Inverting the above—thatis,
∞
[F(ω) exp(−i ω t0 )] = −∞ f (t − t0 )exp(−i ω t)dt—
leads to the result that the Fourier transform for f (t − t0 ) is F(ω) exp(−iω t0 ). (5)
(6) : In order to determine the Fourier transform, F0 (ω), of the function [ f (t) exp
(i ω0 t)] proceed as follows: Write:
∞
F0 (ω) = 1/(2π) [ f (t) exp(i ω0 t)] exp(−i ω t) dt
−∞
∞
= 1/(2π) f (t) exp(−i [ω − ω0 ] t) dt
−∞
= F(ω − ω0 ) . (13.4)
Clearly, therefore, the Fourier transform of [ f (t) exp(i ω0 t)] is F(ω − ω0 ). (6)
13.4 Dirac’s Delta Function 405
13.4 Dirac’s Delta Function
It is necessary to give some description of the function δ(x − a). Indeed, Dirac’s
delta function deserves a detailed and thorough review—see for instance, Dean G.
Du f f y 24. . Still, the following relationships should suffice for the current needs.
For real a
δ(t − a) = ∞ , when x = a
= 0 , when x = a , (13.5)
and
∞
δ(t − a) dt = 1 , (13.6)
−∞
∞
δ(t − a) f (t)dt = f (a) . (13.7)
−∞
A choice popular with electrical engineers is to relate the delta function to the
derivative of Heaviside step function H (t − a). The Heaviside step function is
defined for a ≥ 0 as
H (t − a) = 1, f or t > a
= 0, f or t < a . (13.8)
And its relationship to the delta function is as follows.
d H (t)
δ(t) = . (13.9)
dt
Upon integration, (13.9) yields
t
H (t) = δ(x) dx . (13.10)
−∞
Another useful choice for δ(x) is one recommended by T. B. Boykin 35 . That is
1 1
∞ nπ x
δ(x) = + cos . (13.11)
2L L n=1 L
In order to check whether the Boykin relationship (13.11) is valid, we examine the
accuracy of its following prediction.
406 13 Answer to Additional Assigned Problems
L
f (x) δ(x) dx = f (0) . (13.12)
−L
That is, we check the following relationship for validity:

∞ nπ x
L
1 1
f (x) + cos dx
−L 2L L n=1 L
L ∞ nπ x
L
1 1
= f (x) dx + f (x) cos dx
−L 2L −L L n=1 L
= f (0) . (13.13)
In other words, we examine whether (13.13) indeed holds true—see (13.15) for the
appropriate result for f (0). That is
∞

f (0) = a0 + an . (13.14)
n=1
However, before that can be done the function f (x) needs to be represented in a
suitable format.
To that end, proceed as follows. Given the function f (x) is piecewise differentiable
with period 2 L in the range [-L,L], it can be represented by an infinite Fourier series.
∞
n π x n π x
f (x) = a0 + an cos + bn sin . (13.15)
n=1
L L
The f (x) given above in (13.15) is now inserted into (13.13). Upon working out such
(13.13), we find that the result is the same as that predicted by the use of Boykin’s
equation: Meaning, it is equal to f (0). This fact can be confirmed by comparison
with (13.16).

L ∞
n π x n π x 1
a0 + an cos + bn sin dx
−L n=1
L L 2L

L ∞ n π x n π x 1
∞ nπ x
+ a0 + an cos + bn sin cos dx
−L n=1
L L L n=1 L
∞

= a0 + an = f (0) . Q.E.D (13.16)
n=1
Bibliography
1. Bernouilli, Jacob. (1654)–(1705),

2. Clairaut, Alexis. (1713)–(1765),
3. Lagrange, Joseph - Louis. (1736)–(1813),
4. Riccati, Jacobo, F. (1676)–(1754),
5. Kirchoff, Gustov. (1824)–(1887),
6. Frobenius, Ferdinand, Georg. (1849)–(1917),
7. Euler, Leonard. (1707)–(1783),
8. Runge, Carl, D.T. (1856)–(1927),
9. Kutta, Martin, Wilhelm. (1867)–(1944),
10. Piaggio, H. T. H. “An Elementary Treatise on Differential Equations and Their Applications”.
First published in February, (1920)—nearly a century ago!—by G. Bell and Sons, LTD., and
last reprinted in (1940),
11. Abell, Martha L. “Differential Equations with Mathematica”, Academic Press, INC., (1993),
12. Bajpai, A. C., Calus, J. M., Hyslop, J. “Ordinary Differential Equations”. John Wiley & Sons,
(1970),
13. Cox, W. “Ordinary Differential Equations ”. Arnold, INC., London (1996),
14. Cronin, Jane. “Differential Equations ”. Marcel Dekker, INC., London (1994),
15. Hagin, Frank G. “First Course in Differential Equations”. Prentice-Hall, INC., (1975),
16. Ince, E. L. “Ordinary Differential Equations”. Dover Publications, INC., (1956), original
(1926),
17. Ross, Shepley L. and Ross, Shepley L., II “Introduction to Ordinary Differential Equations ”.
John Wiley and Sons, (1989),
18. Tierney, John A. “Differential Equations”. Allyn & Bacon, INC., (1979),
19. Waltman, Paul. “A Second Course In Elementary Differential Equations”. Academic Press
INC., (1986),
20. Wylie, C. Ray. “Differential Equations”. John Wiley INC., (1979),
21. Yates, Robert. C. “Differential Equations”. McGRAW-HILL INC., (1952),
22. Boas, Mary. L. “Mathematical Methods in Physical Sciences”. John Wiley and Sons INC.,
(2006),
23. Ohm, Georg Simon. (1789)–(1854),
24. Duffy, Dean. J. “Green’s Functions with Applications”. Second Edition, Chapman Hall, CRC
Press., (2015),
25. Laplace, Pierre-Simon. (1749)–(1827),

https://doi.org/10.1007/978-3-319-76406-1
408 Bibliography
26. Green, George. “Essay on the Application of Mathematical Analysis to the Theory of Electricity
and Magnetism”. (1793)–(1841),
27. Maclaurin, Colin. (1698)–(1746),
28. Taylor, Brook. (1685)–(1731),
29. Hook, R. (1635)–(1703),
30. Whittaker, E. T. and Watson, G. N. “Modern Analysis”, pp.194–203, Macmillan, N.Y. (1943),
31. Ampere, Andre-Marie. (1775)–(1836),
32. Volta, Alessandro, G. A. A. (1745)–(1827),
33. Dirac, Paul, Adrien, Maurice. (1902)–(1984),
34. Young, Peter. November (2009), physics.ucsc.edu.
35. Boykin, Timothy, B. Am. J. of Phys. 71, 462–468 (2003)

Raza Tahir-Kheli - Ordinary Differential Equations - Mathematical Tools For Physicists-Springer International Publishing (2018)

Uploaded by

Copyright:

Available Formats

You might also like

Raza Tahir-Kheli - Ordinary Differential Equations - Mathematical Tools For Physicists-Springer International Publishing (2018)

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Raza Tahir-Kheli - Ordinary Differential Equations - Mathematical Tools For Physicists-Springer International Publishing (2018)

Uploaded by

Copyright:

Available Formats

ISBN 978-3-319-76405-4 ISBN 978-3-319-76406-1 (eBook)

Library of Congress Control Number: 2018933457

© Springer Nature Switzerland AG 2018

This book is intended as a reader-friendly source for self-study and as an accessible

numbered in seriatim. Included in the numbering process is the chapter number. As

Philadelphia, USA Raza Tahir-Kheli

4.3 Bernouilli Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

6.1.5 Informal Solution . . . . . . . . . . . . . . . . . . . . . . . . . . 124

6.9.4 Problems Group IX . . . . . . . . . . . . . . . . . . . . . . . . . 167

7.5 Other Special Situations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216

9.1.6 Solution (B) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269

10.2.1 Runge–Kutta Steps . . . . . . . . . . . . . . . . . . . . . . . . . 311

11.9.1 Second Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . 362

12.23 Problems Group I, 11-chapt . . . . . . . . . . . . . . . . . . . . . . . . . . 396

All differential equations make use of the differential operator D. A convenient

1.1.1 Laws of Addition

dν1 f (x) dν2 f (x) dν2 f (x) dν1 f (x)

as well as the distributive law of addition–multiplication.

Also the associative law of addition is obeyed.

© Springer Nature Switzerland AG 2018 1

1.1.2 Laws of Multiplication

Similarly, both the associative law,

and the commutative law,

1.1.3 What is D−1 f (x)?

leads to its solution

where σ0 is a constant. If we assume D −1 represents integration with respect to x,

Another way of looking at this is to write

D −1 . D y = D −1+1 y = D 0 y(x) = y(x) . (1.8)

On the other hand, D ν . D −ν x μ can be written as

1.1.4 Index Law

In (1.7)–(1.11), we have made use of the index law that asserts:

D −1+3 x 2 = D −1 . [D 3 x 2 ] = D −1 . [0] = σ1 . (1.14)

Thus, an additional instruction would be needed to set σ1 = 2 . To sum up:

When n is a positive integer, D n may be treated as an ordinary algebraic

A wealth of terminology is used in the literature. Unfortunately, some sounds similar

2.1 Ordinary Differential Equation

A function involving a dependent variable, say u, only a single independent variable,

An (ODE) may be expressed in the following explicit form

Fx pl (x, u, u (1) , . . . , u (ν−1) ) = u (ν) . (2.1)

and Fx pl is a continuous function of the variables x, u, u (1) , . . . , u (ν−1) , where the

© Springer Nature Switzerland AG 2018 5

Such an equation is called an explicit ordinary differential equation. Knowledge of

is equivalent to having an explicit solution of the explicit ordinary differential equa-

When a given (ODE) cannot be expressed in the explicit from, it is said to be an

Fimpl (x, u, u (1) , . . . , u (ν−1) , u (ν) ) = 0. (2.4)

[Unfortunately, this terminology is somewhat confusing because to many readers the

f (λx, λy, λz) ≡ λn f (x, y, z). (2.7)

F1 (λx, λy) dx + F2 (λx, λy) dy = λn [F1 (x, y) dx + F2 (x, y) dy , (2.8)

2.1.7 Partial Differential Equation

2.1.8 The Order of an (ODE)

2.1.9 The Degree of an (ODE)

2.1.10 Order and Degree: Exercises

Determine the order and degree of the following differential equations.

The highest order derivative in (2.10) is D 3 u. Therefore, it is of third-order. Also, the

The highest-order derivative is clearly of power unity. Therefore, (ODE) (2.11) is of

Now transfer 1 + 3 D 3 u to the left-hand side